1141950

other2022
CRUXXX10.1177/14736691221141950Journal of the ICRU

ICRU REPORT No. 97

Journal of the ICRU
22(1) 1–100
© 2022, International Commission
on Radiation Units & Measurements.
All rights reserved.
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/14736691221141950
https://doi.org/10.1177/14736691221141950
journals.sagepub.com/home/cru

ICRU REPORT 97: MRI-GUIDED

Radiation Therapy Using MRI-Linear
Accelerators

Paul J. Keall (Chair)1, Carri K. Glide-Hurst (Vice-Chair)2, Minsong Cao3, Percy Lee4,
Brad Murray5, Bas W. Raaymakers6, Alison Tree7, and Uulke A. van der Heide8

1
ACRF Image X Institute, Faculty of Medicine and Health, University of Sydney, Sydney, NSW, Australia
2
University of Wisconsin–Madison, Madison, WI, USA
3
University of California, Los Angeles, Los Angeles, CA, USA
4
MD Anderson Cancer Center, Houston, TX, USA
5
University of Alberta/MagnetTx, Edmonton, AB, Canada
6
University Medical Center Utrecht, Utrecht, The Netherlands
7
Royal Marsden Hospital and Institute of Cancer Research, London, UK
8
Netherlands Cancer Institute, Amsterdam, The Netherlands

In collaboration with the ICRU Commission: Vincent Grégoire (Chair), Soren M. Bentzen, John M. Boone,
François Bochud, María-Ester Brandan, David T. Burns, Roger W. Howell, David A. Jaffray, Thomas R. Mackie, Pawel Olko,
Brian O’Sullivan, Thomas Otto, David W.O. Rogers, Norio Saito, Tzu-Chen Yen
Correspondence:
International Commission on Radiation Units and Measurements, 7910 Woodmont Avenue, Suite 400, Bethesda, MD
20814-3095, USA.
Email: [email protected]

THE INTERNATIONAL COMMISSION ON

RADIATION UNITS AND
MEASUREMENTS
December 2022
2 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

ICRU REPORT 97: MRI-GUIDED RADIATION THERAPY USING

MRI-LINEAR ACCELERATORS

Report Committee

Paul J. Keall (Chair), University of Sydney, Sydney, NSW, Australia

Carri K. Glide-Hurst (Vice-Chair), University of Wisconsin–Madison, Madison, WI, USA
Alison Tree, Royal Marsden Hospital and Institute of Cancer Research, London, UK
Percy Lee, MD Anderson Cancer Center, Houston, TX, USA
Brad Murray, University of Alberta/MagnetTx, Edmonton, AB, Canada
Bas W. Raaymakers, University Medical Center Utrecht, Utrecht, The Netherlands
Uulke A. van der Heide, Netherlands Cancer Institute, Amsterdam, The Netherlands
Minsong Cao, University of California, Los Angeles, Los Angeles, CA, USA

Commission sponsor
David W.O. Rogers, Carleton University, Ottawa, ON, Canada

The Commission wishes to express its appreciation to the individuals involved in the preparation of this
Report for the time and efforts that they devoted to this task and to express its appreciation to the orga-
nizations with which they are affiliated.

All rights reserved. No part of this book may be reproduced, stored in retrieval systems, or transmit-
ted in any form by any means, electronic, electrostatic, magnetic, mechanical photocopying, recording
or otherwise, without the permission in writing from the publishers.
 ICRU 3

The International Commission on Radiation

Units and Measurements

Introduction the widest possible range of applicability. Situations can

arise from time to time for which an expedient solution of a
The International Commission on Radiation Units and current problem is required. The ICRU invites and welcomes
Measurements (ICRU), since its inception in 1925, has had constructive comments and suggestions regarding its recom-
as its principal objective the development of internationally mendations and reports. These may be transmitted to the
acceptable recommendations regarding: Chairman.

1. quantities and units of ionizing radiation and

radioactivity, Current Program
2. procedures suitable for the measurement and applica- The Commission recognizes its obligation to provide guid-
tion of these quantities in clinical radiology and ance and recommendations in the areas of radiation therapy,
radiobiology, and radiation protection, and the compilation of data important to
3. physical data needed in the application of these pro- these fields, and to scientific research and industrial applica-
cedures, the use of which tends to assure uniformity tions of radiation. Increasingly, the Commission is focusing
in reporting. on the problems of protection of the patient and evaluation of
image quality in diagnostic radiology and radiation oncol-
The Commission also considers and makes similar types ogy. These activities do not diminish the ICRU’s commit-
of recommendations for the radiation protection field. In this ment to the provision of a rigorously defined set of quantities
connection, its work is performed in cooperation with the and units useful in a very broad range of scientific endeavors.
International Commission on Radiological Protection The Commission is currently engaged in the formulation of
(ICRP). ICRU endeavors to collect and evaluate the latest ICRU Reports treating the following subjects:
data and information pertinent to the problems of radiation
measurement and dosimetry and to recommend the most Dose Prescription, Reporting and Recording in Advanced
acceptable numerical values for physical reference data and Optimization Strategies: Application to Dose Painting
techniques for current use. The Commission’s recommenda- and Robust Planning
tions are kept under continual review to keep abreast of the ICRU Composite Glossary
rapidly expanding uses of radiation. The ICRU feels that it is Stochastic Nature of Radiation Interactions:
the responsibility of national organizations to introduce their Microdosimetry
own detailed technical procedures for the development and Quantitative Imaging for Assessment of Response in
maintenance of standards. However, it urges that all coun- Oncology
tries adhere as closely as possible to the internationally rec-
ommended basic concepts of radiation quantities and units. The Commission continually reviews progress in radia-
The Commission maintains and develops a system of quanti- tion science with the aim of identifying areas in which the
ties and units and concepts (e.g., for radiation therapy) and development of guidance and recommendations can make an
guidance for measurement procedures and techniques having important contribution.
4 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

ICRU’s Relationship with Other International Labour Office

Organizations International Organization for Medical Physics
International Radiation Protection Association
In addition to its close relationship with the ICRP, ICRU has International Union of Pure and Applied Physics
developed relationships with national and international agen- United Nations Educational, Scientific and Cultural
cies and organizations. In these relationships, ICRU is looked Organization
to for primary guidance in matters relating to quantities,
units, and measurements for ionizing radiation, and their The Commission has found its relationship with all of these
applications in the radiological sciences. In 1960, through a organizations fruitful and of substantial benefit to the ICRU
special liaison agreement, ICRU entered into consultative program.
status with the International Atomic Energy Agency (IAEA).
The Commission has a formal relationship with the United
Nations Scientific Committee on the Effects of Atomic Operating Funds
Radiation (UNSCEAR), whereby ICRU observers are Financial support has been received from the following
invited to attend annual UNSCEAR meetings. The organizations:
Commission and the International Organization for
Standardization (ISO) informally exchange notifications of American Association of Physicists in Medicine (AAPM)
meetings, and ICRU is formally designated for liaison with International Atomic Energy Agency (IAEA)
two of the ISO technical committees. The ICRU is a member International Radiation Protection Association (IRPA)
of Consultative Committee for Units (CCU)—a liaison of Radiological Society of North America (RSNA)
Consultative Committee for Ionizing Radiation (CCRI)—
and a member of its sections CCRI(I), CCRI (II), and CCRI In addition to the direct monetary support provided by
(III). The ICRU also enjoys a strong relationship with its sis- these organizations, many organizations provide indirect
ter organization, the National Council on Radiation support for the Commission’s program. This support is pro-
Protection and Measurements (NCRP). In essence, ICRU vided in many forms, including, among others, subsidies for
and NCRP were founded concurrently by the same individu- (1) the time of individuals participating in ICRU activities,
als. Presently, this long-standing relationship is formally (2) travel costs involved in ICRU meetings, and (3) meeting
acknowledged by a special liaison agreement. The ICRU facilities and services. In recognition of the fact that its work
also exchanges reports with the following organizations: is made possible by the generous support provided by all the
organizations supporting its program, the Commission
Bureau International des Poids et Mesures expresses its deep appreciation.
European Commission Vincent Grégoire
International Council for Science Chair, ICRU
International Electrotechnical Commission Lyon, France
 ICRU 5

MRI-GUIDED RADIATION THERAPY USING

MRI-LINEAR ACCELERATORS

Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Glossary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1 Rationale for this Report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 Scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Introduction to MRIgRT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2. Promising Applications for MRIgRT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1 Limitations of Existing IGRT Strategies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Potential Clinical Advantages of MRIgRT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Limitations of MRIgRT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 Existing Clinical Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.5 Genitourinary Malignancies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6 Hepatobiliary and Pancreatic Malignancies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7 Oligometastases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.8 Breast Cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.9 Other Clinical Indications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.10 Immobilization, Patient Setup, Repositioning, Patient Comfort, Claustrophobia . . . . . . . . . . . . . . . . . . . . . . . 24
3. MRI-Linac Interoperability and Existing MRI-Linac Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1 Elekta—Unity Design Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 ViewRay—MRIdian Design Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 MagnetTx Aurora-RT Design Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Australian MRI-linac Design Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4. MRIgRT Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1 Treatment Planning System Commissioning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Absorbed-Dose Calculation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 Treatment Planning Strategies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.4 Deformable Image Registration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.5 MRI-only Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.6 When can we Safely Reduce the PTV Margin? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5. Imaging with an MRI-Linac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1 Image Quality for MRI Guidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Technical Requirements of Sequences for MRI Guidance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3 Mitigation of Susceptibility Artifacts Due to Metallic Implants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4 Sequences for Response Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6. Online Adaptive MRIgRT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.2 Existing Implementations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.3 Workflow and Technical Considerations for Online MRIgRT Adaptive Radiotherapy. . . . . . . . . . . . . . . . . . . . 47
6.3.1 Online Adaptive Treatment Planning Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.4 Quality Assurance for MRIgRT Adaptive Radiotherapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.5 Challenges for MRIgRT ART. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7. MRIgRT Staffing, Training and Safety. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.1 Staff Requirements to Start and Maintain a Program. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.2 Training. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

7.3 MRI Screening. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

7.4 Radiation Safety. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8. MRIgRT Commissioning and Dosimetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8.1 MRI-Linac Commissioning and Ongoing QA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8.2 Reference Dosimetry: Beam Quality Specification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.3 Dosimetry Equipment and Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.4 Impact on Ionization Chamber Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.5 Phantoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.6 Motion/Adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.7 Surface and Exit Dosimetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.8 End-to-End Testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
9. Summary, Recommendations and Future Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.2 Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.2.1 New MRIgRT Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
9.2.2 Experienced MRIgRT Adopters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2.3 Hospital Management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2.4 Industry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2.5 Government. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2.6 Additional Stakeholders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.3. Future Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.3.1 Health Care Resource Changes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.3.2 Clinical Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.3.3 Technical Advances Needed to Further Improve Patient Outcomes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10. Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Appendix A: Table of Resources that Augment this ICRU Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Appendix B: Safety Screening Form Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

B.1 Initial MRI-Linac Safety Screening Questionnaire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
B.2 Daily MRI-Linac Safety Screening Questionnaire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Appendix C: Clinical Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.1 Prostate Cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.1.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.1.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
C.1.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.8 Treatment Delivery and Daily Adaptation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
C.1.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2. Oligometastatic Progression in the Upper Abdomen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
C.2.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
C.2.8 Treatment Delivery and Daily Adaptation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
C.2.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
 ICRU 7

C.3. Recurrent Oligometastatic Disease in the Pelvis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

C.3.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.6 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
C.3.7 Treatment Delivery and Daily Adaptation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
C.3.8 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.4. Partial Breast Radiotherapy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.4.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.4.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.4.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.4.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4.8 Treatment Delivery and Daily Adaptation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.4.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
C.5. Multi-target Oligometastatic Case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
C.5.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
C.5.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
C.5.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
C.5.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
C.5.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
C.5.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.5.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.5.8 Treatment Delivery and Adaptation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.5.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.6. Head and Neck Cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.6.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
C.6.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
C.6.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
C.6.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
C.6.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
C.6.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
C.6.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
C.6.8 Treatment Delivery, Serial Adaptation, and Quantitative Imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
C.6.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
C.7. Ventricular Tachycardia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
C.7.1 Clinical Scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
C.7.2 Treatment Intent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
C.7.3 Patient Positioning and Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
C.7.4 Target Volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.7.5 Organs at Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.7.6 Planning Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.7.7 Treatment Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.7.8 Treatment Delivery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
C.7.9 Treatment Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Preface

The International Commission on Radiation Units and and conventional X-ray image–guided radiation therapy
Measurements (ICRU) has been providing guidance for (IGRT) is shown in the Preface Fig. An additional advan-
more than 90 years to the fields of radiation science, radia- tage of MRIgRT over conventional X-ray IGRT is that
tion medicine, and radiation protection. One of the principal MRIgRT does not add X-ray imaging dose (Ding et al.,
objectives of the ICRU is the development of internationally 2018).
accepted recommendations regarding procedures suitable The report narrative begins with an overall rationale and sta-
for the measurement and application of radiation in medi- tus of MRIgRT, followed by the clinical indications for patient
cine. Aligned with this objective, this ICRU report on MRIgRT treatments across different cancer sites. A summary
Magnetic Resonance Imaging (MRI)–guided Radiation of the MRI-linac interoperability challenges and an overview
Therapy (MRIgRT) using MRI-linear accelerators (linacs) of the MRIgRT systems that have been developed to date are
was commissioned. given. Then, the MRIgRT process steps are given chronologi-
To write this report, a team with diverse country of cally: treatment planning, image guidance, treatment adapta-
origin, gender, and expertise was assembled, noting that tion, and treatment delivery. Given the differences in the
most of the MRIgRT development had been performed in MRIgRT processes compared with X-ray image guidance, the
high-income countries. The focus of the team was on the staff requirements, training and safety are discussed. The final
differences in the radiation therapy process that the addi- technical section reviews MRIgRT commissioning and
tion of MRI into the treatment room brings to the patient dosimetry.
and to the treatment team. With the number of clinical The recommendations span multiple MRIgRT stake-
systems growing rapidly, and the increase in feature holders: hospital managers, newer and expert users, indus-
availability and automation, the authors endeavored to try and government. A future outlook and future needs of
focus on the top-level concepts that will likely remain where MRIgRT may fit into the health care system of the
relevant amid the ongoing technological changes that future, along with anticipated technological advances, are
accompany fast-emerging technologies. These top-level outlined.
concepts include personalization of imaging, high-qual- A summary of additional resources to augment this
ity intratreatment imaging, routine online adaptive radio- report is briefly described. Finally, to give new users real-
therapy, integrated intratreatment imaging with beam world examples of MRIgRT in practice, some clinical case
gating, and integrated functional imaging capabilities. A studies for a variety of sites and treatment complexity are
summary of the key feature differences between MRIgRT given.
 ICRU 9

Preface Figure. An illustration, using example liver cancer images, of some key differences between MRI-guided and conventional
X-ray–guided linear accelerator radiation therapy. The routinely available quality of the anatomic images and the growing availability of
functional images are driving the MRI-linear accelerator uptake. Adapted from Keall et al. (2022).
10 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Glossary

The Glossary of terms used in this report is given below. (2010) and ICRU 96 (Sgouros et al., 2021), we have endeav-
Where similar terms are used to those in ICRU Report 83 ored to be consistent.

Acronym or
Term symbol Definition
Absorbed dose D The mean value of the energy imparted per mass, d ε−/ dm, by ionizing
radiation. The unit of absorbed dose is joule per kilogram, with the special
name gray (Gy) when applied to ionizing radiation.
Accelerated partial breast APBI A hypofractionated treatment in which the tumor cavity and surrounding
irradiation tissue are treated rather than the whole breast.
Adaptive radiation ART A radiotherapy technique that monitors the patient’s anatomy and/or
therapy physiology and changes the treatment plan during the course of treatment
to improve delivery precision and patient outcomes (Hewson et al., 2020).
ART typically involves temporally changing the treatment plan in response
to treatment-induced changes, anatomical motion, or functional variations.
ART is often defined by timescales: offline between fractions, online
immediately before a fraction, and in real-time during a fraction as further
explained in Table 1.1.
Adapt to position ATP To adjust beams to the imaged target position, also called a virtual couch
shift as the couch cannot be moved in three-dimensions. In the ATP
process, no daily structure delineation is performed, and translation-only
rigid registration is implemented to determine the isocenter shift.
Adapt to shape ATS A pretreatment adaptive process that consists of deformable image registration
and contour modification to generate a new adaptive plan of the day.
American Association of AAPM TG The American Association of Physicists in Medicine is a scientific,
Physicists in Medicine educational, and professional organization of Medical Physicists. Task
Task Group Groups are formed to develop consensus guidelines and reports.
American College of ACR A professional medical society representing diagnostic radiologists, radiation
Radiology oncologists, interventional radiologists, nuclear medicine physicians, and
medical physicists.
Apparent diffusion ADC Parameter calculated using diffusion-weighted imaging to quantify the
coefficient magnitude of diffusion of water molecules within tissue.
Balanced steady-state b-SSFP MRI rapid gradient-echo sequences commonly used for noncontrast
free precession scanning in the chest, abdomen, and pelvis.
Cone beam computed CBCT Volumetric imaging created by reconstructing kilovoltage X-ray projections
tomography acquired from multiple views of the object.
Contrast-to-noise ratio CNR Ratio of the absolute difference in intensities between two regions to the
level of fluctuations in intensity due to noise.
Common terminology CTCAE Common Terminology Criteria for Adverse Events is the standard
criteria for adverse classification and severity grading scale for adverse events in cancer
events therapy, clinical trials, and other oncology settings.
 ICRU 11

Acronym or
Term symbol Definition
Computed tomography CT A computerized X-ray imaging procedure in which a beam of X-rays is
aimed at a patient and quickly rotated around the body. The transmitted
fraction of photons that pass through the patient are measured and used to
generate cross-sectional images (“slices”) of the body.
Clinical target volume CTV A volume of tissue that contains the gross tumor volume and/or subclinical
malignant disease with a certain probability of occurrence considered
relevant for therapy.
Deformable image DIR A spatially variant registration transformation yielding a unique
registration displacement vector for every voxel in the source data set as applied for
contour propagation, dose accumulation, and image-to-image mapping.
Deformation vector field DVF The result of a nonrigid registration between two images.
Electron return effect ERE The impact of the Lorentz force on charged particles at interfaces of different
density.
Electron streaming effect ESE The out-of-field surface dose contribution due to air-generated or patient-
ejected electrons affected by the Lorentz force.
Field of view FOV The region superimposed over the human body over which MRI data are
acquired.
Flattening filter free FFF Linear accelerator configuration where the flattening filter is not present.
Without a flattening filter, the beam has higher output, lower mean energy,
larger intensity variations but less beam spectrum changes across the
treatment field, and less scattering outside the field. FFF beams should only
be used with IMRT delivery.
Gradient nonlinearity GNL Variations in the linearity of the applied magnetic field gradients that can
lead to geometric distortion and other issues in reconstructed images.
Gray Gy SI unit of absorbed dose corresponding to 1 J/kg.
Gross tumor volume GTV The gross demonstrable extent and location of the tumor.
Image-guided radiation IGRT The use of in-treatment room imaging to improve the alignment between the
therapy radiation beam and the treatment target, accounting for interfraction and/or
intrafraction changes.
International ICRU An international organization that commissioned this report. The ICRU has
Commission on provided guidance for more than 90 years to the fields of radiation science,
Radiation Units and radiation medicine and radiation protection.
Measurements
Intensity-modulated IMRT A type of three-dimensional external beam radiation therapy that uses
radiation therapy computer-generated images to conform radiation delivery to the size and
shape of a tumor. In IMRT, thousands of tiny radiation beamlets enter the
body from many angles and intersect the tumor. As the intensity of each
beamlet can be controlled, the spatial distribution of absorbed dose can be
shaped to conform to the tumor. The aim is to deliver a more conformal
radiation dose to a tumor while reducing the dose to nearby healthy tissue.
Linear accelerator linac A device that accelerates electrons to high energy to create electron or
photon beams, with one application being their use in radiation therapy.
Magnetic field strength B0 The magnitude (SI unit, tesla) of the magnetic field strength of the MRI
scanner. More generally, B0 is the net magnetization vector and as such
has both magnitude and direction. In MRI, the direction of B0 is along the
center of the bore of the magnet for closed bore systems and aligned with
the center of the poles for open bore systems.
12 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Acronym or
Term symbol Definition
Magnetic resonance MR Magnetic resonance is a quantum mechanical resonant effect that can appear
when a magnetic dipole is exposed to a static magnetic field and perturbed
with another, oscillating electromagnetic field.
Magnetic resonance MRI Use of magnetic resonance to create images of objects such as the body.
imaging Currently, this primarily involves imaging the distribution of mobile
hydrogen nuclei (protons) in the body. The image brightness depends
jointly on the spin density and the relaxation times.
MRI-guided Radiation MRIgRT In this report MRIgRT is restricted to MRI-linacs.
Therapy
MRI-linac A device for MRIgRT that integrates an MRI scanner and linear accelerator.
Monitor units MU A measure of machine output from a medical accelerator used in radiation
therapy
Multileaf collimator MLC A radiation beam shaping device that is made of individual “leaves” of a
high density and atomic number material, typically tungsten. The leaves
can move independently in and out of the path of the radiotherapy beam to
shape and vary the beam intensity.
MLC tracking The process of accounting for motion during radiotherapy by using a real-
time position monitoring system and adjusting the radiation beam using the
MLC, and in some cases the beam jaws, to improve the alignment between
the beam and the target. The goal is to improve the agreement between the
planned and delivered dose distributions to the treatment target or targets
and OARs.
Organ at risk OAR An organ at risk has normal tissues whose radiation sensitivity may
significantly influence treatment planning and/or prescribed dose (e.g.,
spinal cord). The dose-volume response of normal tissues is a complex
process, which changes progressively. Used in external beam radiation
therapy planning. OARs are organs at risk.
Planning organ at risk PRV A geometrical concept to ensure sparing of an OAR can be achieved with
volume a certain probability by the expansion of the OAR volume. An OAR
expansion to a PRV is analogous to the CTV expansion to the PTV.
Planning target volume PTV A geometrical concept introduced for treatment planning and evaluation. It
is the recommended tool to shape absorbed-dose distributions to ensure
that the prescribed absorbed dose will actually be delivered to all parts
of the CTV with a clinically acceptable probability, despite geometrical
uncertainties such as organ motion and setup variations.
Positron emission PET Nuclear medicine three-dimensional imaging technique that uses annihilation
tomography photons.
Quality assurance QA Procedures and checks to reduce variation in a process.
Quality control QC Procedures and measurements to monitor variation in a process.
Quantitative imaging QIB An objective characteristic derived from an in vivo image that can be
biomarker measured and is an indicator of normal biological processes, pathogenic
processes or a response to a therapeutic intervention.
Radiation therapist — A member of the medical team who administers doses of radiation to
patients with cancer or other diseases. The training and roles of radiation
therapists vary by country. Also, in different countries a radiation therapist
is referred to as a radiation technologist, radiation therapy technologist,
radiologic technologist, or therapeutic radiographer. The term radiation
therapist is used in this report.
Radiation therapy RT The use of high-intensity ionizing radiation to treat cancer and other diseases.
 ICRU 13

Acronym or
Term symbol Definition
Radiofrequency RF The electromagnetic waves used to excite the hydrogen nuclei in MRI.
Real-time The latency or response time of the motion-tracking system is less than the
timescale of significant changes in the target position.
Specific absorption rate SAR RF energy during MRI scans is deposited mostly in the form of heat. The
SAR is the energy dissipated in tissue (watts) per kilogram of tissue mass.
Stereotactic ablative SABR A high precision radiation therapy technique that is used to deliver a much
radiation therapy higher dose of radiation to the tumor in fewer treatment fractions than
conventional radiation therapy. The SABR treatments typically involve
between one and five treatments, meaning fewer patient visits to the
hospital. Synonymous with SBRT.
Stereotactic body SBRT Synonymous with SABR.
radiation therapy
Source axis distance SAD The distance between the source of ionizing radiation and the axis of rotation
of the treatment gantry.
Signal-to-noise ratio SNR The relative contributions of the true signal and random superimposed
signals (“noise”) to a detected signal.
Standard operating SOP A set of written instructions that describes the step-by-step process that must
procedure be taken to properly perform a routine.
Source surface distance SSD The distance between the source of the ionizing radiation and the surface of
the patient or phantom.
Turbo field echo TFE A gradient echo pulse free precession sequence used for achieving rapid and
high quality imaging
Tissue phantom ratio TPRx,y The ratio of absorbed dose at a given depth, x, to the absorbed dose at fixed
reference depth, y, in water phantoms when the point of measurement is at
the same distance from the source and the same collimator setting is used.
For example, TPR20,10 is the ratio of the absorbed dose at 20 cm to that at
10 cm depth for a 10x10 cm2 field size at the measurement depth when
used as a beam quality specifier.
Treatment planning TPS A computer system used to determine a variety of parameters including
system beam arrangements, energies, and field sizes, to produce a safe and
effective absorbed-dose distribution for radiation therapy.
14 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

1. Introduction

1.1 Rationale for This Report To focus the clinical relevance at the time of publishing
the report, 60Co MRIgRT and particle therapy MRIgRT are
MRI-guided Radiation Therapy (MRIgRT) using integrated not included in the scope of this report. However, many of
MRI-Linear Accelerators, or MRI-linacs, is a rapidly advanc- the principles outlined in this report will apply to these tech-
ing technology implemented in radiation therapy, offering the nologies. The 60Co MRIgRT systems were used to develop
ability to visualize soft tissue and to adapt the treatment to and clinically pioneer MRIgRT and provide much of the
anatomical and physiological changes observed prior to and early outcome data, although they are being phased out of the
during radiation therapy. Acquiring detailed, high-contrast market. Particle MRIgRT systems have been the subject of
magnetic resonance (MR) images in the treatment position research studies (Oborn et al., 2017) and prototype systems
both before and during treatment are new information sources are under development (Schellhammer et al., 2018), although
and workflow changes that are specific to MRI-linacs. For the broad clinical implementation of these systems is not
head and neck cancer patient treatments, breathing and swal- anticipated in the immediate future.
lowing that is occurring during treatment can be observed in
exquisite detail. For lung cancer patients, the breathing lungs
1.3 Introduction to MRIgRT
and beating heart are seen during treatment. Compared with
standard-of-care X-ray imaging–based radiation therapy With advances in engineering, computing, imaging, and arti-
technologies, MRIgRT has the potential for improved tumor ficial intelligence, the ability of the radiation therapy process
control and decreased normal tissue complications obtained to selectively target cancerous tissue and spare healthy tissue
by reducing the safety margins associated with positional is increasing in the radiation therapy process. One of the
uncertainty. This, in turn, is likely to increase the number of major innovations to emerge in radiation therapy over the
patients eligible for shorter treatment courses, thereby bene- past decade is MRIgRT which uses MRI as the primary
fiting patients. New and emerging technology is being devel- localization and treatment planning modality while also
oped for imaging, absorbed-dose delivery, and absorbed-dose being acquired during the delivery of radiation therapy to
measurement. In parallel, this promising technology evalu- visualize the internal anatomy and location of the patient.
ated in clinical research studies has highlighted differences in These images may be used to adapt the treatment to account
acute normal tissue toxicities and tumor control compared for the anatomic changes observed since the original treat-
with X-ray imaging–based radiation therapy (Kishan et al., ment plan was generated. The enhanced visualization of tar-
2022b; Ma et al., 2022). The rationale for this report is to give gets and organs at risk (OARs) has improved the precision of
the radiation therapy community guidance on the safe and irradiation and thereby enabled ultra-hypofractionation.
effective clinical implementation of MRIgRT and highlight Daily imaging has improved response assessment, and in
emerging evidence supporting it use. Supplemental docu- particular advanced imaging techniques, such as diffusion-
ments to read that provide background and support for this weighted MRI, allow daily monitoring of the changing bio-
report are given in Table A.1 in Appendix A. logical characteristics of the target. After treatment, daily
dose may be reconstructed, and dose accumulation per-
formed to assess the overall dose to the patient during the
1.2 Scope treatment course.
This report covers MRIgRT using integrated MRI-linacs. With this functionality, the introduction of MRIgRT is not
Many aspects of this promising emerging technology are an incremental improvement of radiotherapy. Rather, it has
addressed to maximize the safety and clinical benefit to be the potential to change the paradigm of radiation oncology.
achieved by MRIgRT. As MRIgRT involves the merging of Two concepts that were introduced decades ago are now
both magnetic resonance imaging (MRI) and radiation ther- brought together (van Houdt et al., 2021b). The concept of
apy (RT), increased complexity exists that introduces new adaptive radiotherapy was introduced by Yan et al. in 1997.
risks and requires learning new information for those who It creates a closed-loop process where the treatment plan can
may only have experience in one of these fields. be modified based on observed changes in the patient
 Introduction 15

Figure 1.1. Key elements of a generalized MRIgRT process. The actual steps taken for each treatment will depend on the patient and
disease features, the treatment intent, and the technology available. Workflows with and without adaptive and with and without preplanning
are given in the inset. The workflow for a nonadaptive preplanned MRIgRT treatment progresses through all of the steps apart from the
plan in step 6; an adaptive preplanned MRIgRT treatment progresses through all of the steps. When planning is performed on the MRI-linear
accelerator, steps 1 to 3 are incorporated into the MRIgRT suite workflow. QA = quality assurance.

(Jaffray, 2012; Yan et al., 1997). Online X-ray adaptive Quality Control (QC) instruments for measuring absorbed
radiotherapy was clinically pioneered using CT-on-rails sys- dose.
tems (Schwartz et al., 2012) and is now commercially avail- MRI is a unique and distinct imaging modality. MRI is
able using CBCT guidance (Sibolt et al., 2021). The concept newer to radiation therapy than X-ray imaging, which has
of dose painting was introduced in 2000 (Ling et al., 2000). long been the mainstay of image-guided radiation therapy.
The idea is to improve the therapeutic window by matching The complexity and lack of familiarity leave a training gap
the absorbed-dose distribution to the biological heterogene- that needs to be addressed for all members of the treatment
ity as observed with functional imaging such as identifying a team, including medical physicists, nurses, radiation oncolo-
dominant intraprostatic lesion in the prostate on a daily basis gists, and radiation therapists.
with MRIgRT (Almansour et al., 2021). With MRIgRT, the To integrate a linear accelerator with an MRI scanner,
enhanced visualization of targets and OARs and the capacity modifications have to be made to the MRI system and the
for daily response assessment create the potential for adap- linac. In the design of the commercially available systems,
tive radiotherapy that considers both morphological and bio- different choices have been made. The MRIdian system of
logical changes. This concept is being investigated in ViewRay Inc (Cleveland, Ohio) uses a dedicated design
ongoing clinical trials (Bahig et al., 2018). Indeed, with the based on a 0.35 T magnet. The Unity system of Elekta
soft tissue contrast and associated lesion conspicuity afforded (Stockholm, Sweden) has modified a diagnostic 1.5 T scan-
by MRIgRT, margin reduction has been realized across sev- ner. To create a window for irradiation, the gradient coils
eral disease sites, including prostate (Kishan et al., 2022b; have been split. Both the MRIdian and Unity systems use
Tetar et al., 2019), rectal (Intven et al., 2021), and treatment radiofrequency (RF) receive coil systems that differ from
of partial breast cancers (Bahig et al., 2018). This margin those applied in diagnostic MR scanners such that they have
reduction coupled with real-time intrafraction motion moni- a low amount of attenuation and are flexible for use in con-
toring may lead to reduced treatment-related toxicities. junction with or suspended over immobilization devices.
In terms of safety, MRI does not add additional absorbed Given the complexity of the hybrid systems, MRI-linacs
dose from ionizing radiation to the patient, thereby reducing have higher capital, operating, staff and workflow costs than
the associated concerns arising from the use of X-ray imag- conventional X-ray-guided linacs.
ing for more frequent imaging and intrafraction monitoring. The MRI magnetic field affects the radiation beam gen-
However, when combining MRI with radiation therapy, the eration and transport to and inside the patient requiring addi-
safety risks of both systems to patients, staff, and visitors are tional considerations during treatment planning. In addition,
higher than for the individual systems. The contraindications instruments and procedures used for QA should be revised
of patients to MRI also apply to a treatment with MRIgRT. for operation in a magnetic field. An overview of challenges
The MRI environment dictates which devices can be brought is described in Hehakaya et al. (2020).
into the treatment space such as Quality Assurance (QA) and Key elements of the MRIgRT process are shown in Fig. 1.1.
16 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Table 1.1. ART timescales. Adapted from Glide-Hurst et al. (2021a).

ART implementation
(timescale) Definition and explanation
Offline Offline ART mostly addresses systematic and progressive changes that occur during the treatment course
(Days) such as patient weight loss and tumor morphological changes. For offline ART, adjustments to a patient’s
treatment plan parameters based on these observed changes are modified after the current treatment
fraction, typically following the same clinical workflow as regular initial treatment planning.
Online Online ART is a process where the patient’s treatment plan is adjusted prior to the treatment delivery to
(Minutes) account for temporal and stochastic changes detected prior to a treatment fraction while the patient remains
in the treatment position. As a result, online ART requires imaging, rapid replanning, plan review, and an
acceptable form of patient-specific quality assurance.
Real-time Real-time ART is where the treatment plan and delivery are automatically adapted to the changing patient
(Seconds) anatomy during treatment delivery without operator intervention. Real-time ART may be performed through
treatment gating, dynamic tracking by the treatment machine (e.g., CyberKnife or Vero [Hiraoka et al.,
2012] systems), by the multileaf collimators, and intrafraction replanning although such an approach typically
requires continuous imaging with constant replanning and rapid absorbed dose calculation (Klawikowski et al.,
2018; Kontaxis et al., 2017; Pathmanathan et al., 2018).
Note. The changes that are adapted to can be anatomic changes and/or functional/physiological changes. ART = Adaptive radiation therapy.

The conceptual framework for MRIgRT builds on previ- Incorporated, Sunnyvale, California) dynamically tracks
ous ICRU reports, particularly ICRU Report 83 (2010). tumors during the respiratory cycle via an external to inter-
Additional useful reports and guideline documents that nal motion correlation model updated throughout treatment
augment this ICRU report are listed in Table A.1 in using X-ray imaging (Hara et al., 2007; Schweikard et al.,
Appendix A. 2004). Similarly, the Radixact (Accuray Incorporated) heli-
Adaptive radiation therapy (ART) can operate over three cal TomoTherapy system has integrated intrafraction
different timescales, offline, online, and real-time. The motion management based on Synchrony to predict motion
choice depends on the rate of change of the anatomy, treat- based on implanted fiducials or the tumor (Ferris et al.,
ment goal, availability of technology and resources, and the 2020). Real-time ART has also been realized using kilo-
overall impact on the treatment workflow as detailed in sec- voltage intrafraction monitoring (KIM) on a conventional
tion 6. As outlined in Table 1.1, the ART implementation is linac that localizes the target based on implanted fiducials
largely governed by the timescale of adaptation. Offline in the prostate and realigns the multileaf collimators
adaptation may be used to change the plan periodically (MLCs) to the target (Keall et al., 2018a). Hybrid
based on the changing tumor volumes, such as in the case approaches are also possible that may incorporate func-
of fractionated chemoradiation therapy for localized and tional MRI information such as employing diffusion-weight
advanced head and neck cancer where bulky lesions and imaging to track tumor response or identify subvolumes of
weight loss may necessitate serial monitoring and plan disease.
updating. One example of weekly offline adaptive therapy The ability to recontour on a daily basis requires the
to account for changes is highlighted in Appendix C. Online multidisciplinary team to think carefully about the impli-
adaptation is used to improve the distribution of absorbed cations of changing the target or OAR contours in response
dose delivered by accounting for the daily anatomical to anatomical variation. The implications of any resulting
changes of the primary tumor, nodes to be treated, and sur- increase or decrease in doses must be considered
rounding OARs. Real-time adaptation is used to shape the (Heukelom and Fuller, 2019). This consideration varies by
radiation beam to optimally account for the respiratory and tumor type and clinical scenario. Five possible scenarios
cardiac-induced motion of the individual targets and OARs. are presented below, which are not necessarily mutually
Currently, various non-MRIgRT approaches have been exclusive:
used for real-time adaptation, including treatment gating
and implementing dynamic tracking by the treatment •• Adaptation to maintain target and OARs dose as spec-
machine (e.g., CyberKnife or Vero [Hiraoka et al., 2012] ified in the pretreatment plan, for example, tumor type
systems). In this setting, intrafraction replanning may be where microscopic disease may persist even as the
implemented using continuous imaging and efficient dose gross tumor volume (GTV) shrinks during treatment.
calculation (Klawikowski et al., 2018; Kontaxis et al., •• Adaptation to shrink GTV contour as tumor regresses
2017). The CyberKnife with Synchrony (Accuray (Bahig et al., 2018).
 Introduction 17

•• Adaptation to maintain target contour and dose but •• Biologically target areas of active or persistent disease
decrease OAR dose by recontouring and re-optimiz- during a course of therapy, for example, dose escala-
ing daily, for example, pancreatic SBRT where duode- tion to areas of persistent disease activity in high-
num moves toward the target (Clinical Trial grade glioma (Clinical Trial NCT04726397).
NCT03621644).
•• Adaptation to escalate target dose when daily OARs When considering any GTV, CTV, or OAR contouring or
anatomy allows (Clinical Trial NCT04595019). recontouring, it is important to understand the underlying
•• Adaptation as a method to manage differential inter- uncertainties (ICRU Report 83, 2010; Weiss and Hess,
fraction motion between multiple targets. An example 2003). There are also variations between MRI and CT
is the case study in section C.5. modalities (Pathmanathan et al., 2019).
18 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

2. Promising Applications for MRIgRT

Adaptive radiotherapy strategies offer the potential to create anatomy is currently only a sufficient method of IGRT when
a plan based on daily anatomy. This will improve the targeted the relationship between the bone, target, and adjacent OARs
absorbed-dose delivery where anatomy may change signifi- is fixed, where large margins are used to compensate for the
cantly from day to day. With conventional image guidance relative motion of bone and target, or in palliative settings.
(i.e., computed tomography [CT]), this is limited by poor
soft tissue discrimination and the penalty of added X-ray
dose. MRIgRT is a natural evolution of image-guided radia-
2.2 Potential Clinical Advantages of
tion therapy (IGRT) practice over the last decade, as MRI MRIgRT
allows superior soft tissue imaging compared with CT, it Clinically, the drivers for MRI guidance are the need for
allows online tumor motion tracking and avoids the inva- more accurate target delineation and treatment delivery. This
siveness of fiducial implantation. However, much of the evi- is particularly relevant for hypofractionated treatments
dence and future potential of MRIgRT lies not just in the where accurate absorbed-dose delivery is more important
margin reduction provided by the enhanced imaging and gat- and longer treatment times are more tolerable. Superior ana-
ing capabilities but also in the ability to create a daily adap- tomical definition on MRI allows better target definition for
tive plan based on the anatomy of the day or even the anatomy the majority of soft tissue tumors. Intrafraction cine-MRI
of the moment. This represents a fundamental change in (use of fast MRI images to capture motion) allows an added
radiotherapy planning and delivery and changes many layer of security, confirming geometric coverage during dose
aspects of clinical and dosimetric decision-making, as well delivery and allows both beam gating and tracking of targets
as bringing new challenges for quality assurance. that move during beam delivery. Tracking facilitates smaller
margins and optimally accurate absorbed-dose delivery. The
2.1 Limitations of Existing IGRT ability to alter and personalize the plan daily, based on up-to-
date anatomy, is of potential value in most cancers and is
Strategies likely to facilitate further reduction in Planning Target
It is now universally acknowledged that image guidance is Volume (PTV) margins offering augmented protection of
needed for optimal radiotherapy delivery. Some IGRT is OARs and/or target dose escalation.
always better than none (Zelefsky et al., 2012), and daily is For many cancers, there is a theoretical advantage to MRI
better than weekly (de Crevoisier et al., 2018), but despite the guidance and rapid daily replanning. Clinical experience is
range of methods available, no single method is optimal in all early and so demonstrable clinical gains are immature; how-
cases. The CT and cone beam CT (CBCT)-based IGRT meth- ever, it is likely that the evidence will evolve quickly. For
ods are abundant and useful but are limited by the similarity example, the MIRAGE study has shown a reduction in acute
in X-ray attenuation between most tissues, particularly those toxicity with MRI-guided prostate radiotherapy (incorporating
without adjacent air-filled cavities. This restricts anatomical smaller margins), compared with CT guidance (Kishan et al.,
matching for most tumor sites, as the majority of tumors are 2022a; Ma et al., 2021a). In a nonrandomized study, MRI-
better seen on MRI than CT. These methods also involve the guided postprostatectomy radiotherapy appears associated
use of ionizing radiation for guidance purposes only. with lower side effects (Ma et al., 2022). Current international
Matching of fiducials on X-ray images is accurate, repro- databases such as assembled in the MOMENTUM trial
ducible, and allows the ability to track (Keall et al., 2018b) (NCT04075305) will form the evidence base in advance of
but gives no volumetric information on the tumor or OARs, larger, later-phase trials, which are planned (Chuong et al.,
and can cause imaging artifacts that obscure the target. 2021; de Mol van Otterloo et al., 2020; Westerhoff et al., 2022).
Fiducial insertion requires an invasive, and very occasionally
dangerous, procedure to implant the fiducials into or near the
2.3 Limitations of MRIgRT
tumor. With multiple targets in the same organ, it is not prac-
tical to have fiducial markers associated with each target. As commercial MRI-linacs are relatively new, the technol-
Other methods, including in-room ultrasound imaging, are ogy is still evolving. There are some issues that either limit
used rarely but are often cumbersome or poorly validated. current use or require workarounds to solve. At present, the
Although historically quite common, planar imaging of bony current MRI-linacs differ from standard linacs in that they
 Promising Applications for MRIgRT 19

cannot move the couch to account for all six degrees of free- systems, such as MR simulators, may also be used. Having
dom (translations and rotations). At present, the MRIdian single modality workflows via an MRI-primary image of the
couch translates in three dimensions, whereas the Elekta same contrast is advantageous in that it assists in the daily
Unity couch cannot be moved to account for inter- or image registration during IGRT to better match like anatomy
intrafraction shifts. Instead, a “virtual couch shift” (“adapt to and overall image intensities to support deformable image
position”) technique has been developed, so that the plan can registration for adaptation.
be shifted, rather than the patient, which achieves a similar or
better outcome than a traditional couch shift. However, this
2.4 Existing Clinical Data
requires extra time on the couch and varying complexities of
recalculation or re-optimization, albeit with short calculation When seeking to test a new radiotherapy technology, such as
times. MRIgRT, a sequence of preclinical and clinical steps must be
An important consideration for using MRI for primary followed for each proposed indication. These are outlined in
planning and the daily IGRT/replanning image is the need to Fig. 2.1 and follow the principles of the R-IDEAL paradigm
evaluate the image quality, particularly for the presence of (Verkooijen et al., 2017).
motion artifacts, the effect of acquisition in a single breath- Various clinical sites have made different amounts of
hold, blur caused by motion, or image truncation. Relevant progress along this pathway so far. To indicate the focus of
artifacts are addressed in section 5.3. current (2022) clinical trials, Table 2.1 gives examples of the
It is imperative to ensure an adequate field of view (FOV) larger currently recruiting clinical trials of MRIgRT. Many
is obtained for all beams traversing the body in addition to smaller or pilot studies of more novel indications for MRIgRT
having sufficient coverage to accommodate dose fall off. are also recruiting and available clinical data will evolve rap-
Respiratory motion can cause artifacts and strategies to miti- idly over the next few years.
gate these are described in more detail in section 5.3. Below, and in the example cases described in Appendix
At present, the added complexity of introducing online C, we have showcased tumor types or anatomical sites at dif-
adaptive radiation therapy to MRIgRT, as outlined in section ferent stages of the developmental pathway described in Fig.
6, highlights the potential impact on throughput and patient 2.1, ranging from cardiac ablation (section C.7) which is in
time on the treatment table, although these aspects would the early clinical assessment phase to prostate (section C.1)
also be present in an X-ray-guided adaptive workflow as and upper abdomen (section C.2) which have some clinical
well. Higher capital costs and lower throughput than non- outcomes data available. Here we discuss the clinical data
adaptive strategies are current limitations to more wide- guiding current practice (as outlined in Appendix C): and
spread adoption. MRIgRT can be used as image guidance discuss the available evidence and future directions for
only, without adaptation, which would significantly mitigate research, some of which are being addressed currently in the
the time and manpower implications. Most centers, however, trials in Table 2.1. Certain clinical indications have been
will only have one MRIgRT device, which can cause prob- selected based on the volume of data supporting the use of
lems with continuity of treatment in the event of a MRIgRT in this setting.
breakdown.
Multileaf Collimator (MLC) tracking (Keall et al., 2021)
2.5 Genitourinary Malignancies
is an evolving motion management method. Beam gating, a
complementary or stand-alone motion management method, Outcomes for localized prostate cancer are now very good
is available on the MRIdian and Unity systems. with over 90 % 5-year control in some series and less than 5 %
There are also limitations due to the bore size, which patients with grade 2+ bowel symptoms at 2 years (Dearnaley
excludes some larger patients and those with claustrophobia, et al., 2016). Therefore, demonstrating improvements in
although experience so far indicates this is rarely a problem. either cure or reductions in toxicity is going to be challeng-
As with all MRI scanners, patients with certain implantable ing for early disease. Nevertheless, many groups have used
devices are also excluded or more difficult to treat (see sec- localized prostate cancer to gain experience with daily
tion 2.10). Field size limitations, due to technical constraints replanning and MRI guidance, which benefits from the supe-
of the machine, may prevent some longer field treatments, rior visibility of the prostate capsule on MRI. A change in
such as extended nodal irradiation. Patient immobilization rectal size and shape, and/or change in seminal vesicle posi-
devices and patient positioning may need to be altered for tion, can result in geographical miss or relative overdose of
MRIgRT treatments, which may provide problems or, con- the rectum or bowel without daily adaptive replanning.
versely, solutions to previous problems with standard Prostate swelling is now known to be significant with SBRT
machines. schedules (Alexander et al., 2022), further strengthening the
It is anticipated that for some workflows the primary rationale for daily adaptation (Ma et al., 2021b). Examples
planning image will be obtained directly from the MRI-linac, of patients who have had significant dosimetric benefit from
which can limit throughput on the machine. Equivalent MRIgRT have been published (Dunlop et al., 2020).
20 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure 2.1. Schematic showing developmental and clinical testing pipeline for MRIgRT. OARs = organs at risk.

Early experience of MRIgRT includes a proportion of as a radiobiological, perspective in prostate cancer (Hall
patients treated for prostate cancer in most mixed tumor et al., 2022).
series (Bertelsen et al., 2019; Henke et al., 2018; Sahin et al., SBRT in five fractions has been delivered to the prostate
2019). Although many start with moderately hypofraction- with good short-term genitourinary (GU) and gastrointesti-
ated regimens, the resource intensity of MRIgRT makes nal (GI) toxicity profiles (Bruynzeel et al., 2019). Only 5.0 %
extreme hypofractionation attractive from a resource, as well of patients experienced grade 2 GI symptoms within 12
 Table 2.1. Current investigational clinical trials of MRIgRT with over 50 patients (where MRIgRT is the focus of the study).
Leading
Name of trial/acronym center Tumor type NCT number n Phase Primary endpoint

Above diaphragm
MRI-linac Guided Adaptive Fractionated Stereotactic RT for Brain China Brain metastases from NCT04946019 55 2 1-year intracranial progression-free survival
Metastases from non–small cell lung cancer NSCLC
UNITED trial USA Glioblastoma Multiforme NCT04726397 97 2 Recurrence at margin of radiation volume within
1 year
MR-ADAPTOR USA Head and neck NCT03224000 75 2 Locoregional control at 6 months
STAR-LUNG study Denmark Ultra-central lung NCT05354596 69 2 Cumulative Grade 4 toxicity
tumors
LUNG STAAR study USA Central and ultra-central NCT04917224 60 2 Treatment completion and absence of Grade
lung tumors 3-5 toxicity within 1 year
MR-linac Guided Adaptive Radiotherapy for Inoperable China Mediastinal tumors NCT05163509 59 2 Local recurrence-free survival at 2 years
Mediastinal Tumor
Abdomen
MAESTRO trial Germany Liver metastases NCT05027711 90 2 Cumulative treatment-related toxicity
SMART trial USA Pancreas NCT03621644 133 2 GI toxicity within 90 days
MASPAC Germany Pancreas NCT05114213 92 2, Randomized Cumulative pain index
SIRIUS trial Netherlands Metastases from NCT05375708 93 2 2-year progression-free survival
Colorectal cancer
THUNDER2 Italy Rectum NCT04815694 63 2 Pathological complete response
Chemotherapy Combined With High-dose Radiotherapy for Low China Rectum NCT05338866 58 2 3-year progression-free survival
Rectal Cancer Using MR Guided Linear Accelerator
Prostate
SHORTER USA Salvage prostate NCT04422132 134 2, Randomized Late GI toxicity
MIRAGE USA Prostate NCT04384770 179 3, Randomized Acute GU toxicity
SCIMITAR USA Salvage prostate NCT03541850 91 2 5-year biochemical control and toxicity
SMILE Germany Prostate NCT04845503 68 2 Toxicity or discontinuation of therapy
Proseven trial Belgium Prostate with GTV NCT04896801 120 2 Acute toxicity
boost
FORT USA Prostate NCT04984343 136 2, Randomized Change in GI quality of life score
ERECT Netherlands Prostate NCT04861194 70 2 Erectile function at 2 years
A Study of MRI-guided High-dose Radiation Therapy in Prostate USA Prostate NCT04997018 91 2 Biopsy positivity rate at 2 years
Cancer
AFFIRM Netherlands Prostate NCT05373316 95 2 Acute toxicity
EXCALIBUR USA Prostate NCT04915508 102 2 Change in GI quality of life score

Note. NCT = national clinical trials; NSCLC = non–small cell lung cancer; GI = gastrointestinal; GU = genitourinary.

21
22 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

weeks of MRIgRT, and 23.8 % experienced grade 2 GU constraints based on the anatomy of the day, a new adaptive
symptoms. No grade 3 GI toxicity was seen; grade 3 GU plan was generated. If this plan either improved target cover-
toxicity was seen in 0 % and 5.9 % according to the Common age or reduced/eliminated OAR constraint violation, the
Terminology Criteria for Adverse Events (CTCAE) and updated plan was used for treatment. Patients in this study
Radiation Therapy Oncology Group toxicity scales, respec- were stratified into those who received high-dose radiation
tively. The authors suggest that with the low rates of toxicity (described by the authors as BED10 > 70 Gy) versus standard
seen in their trial, rectal spacers and implantable markers are doses (BED10 < 70 Gy). In this analysis, patients who
no longer required. Recently presented data from the phase received high-dose radiation, enabled by adaptive MRIgRT,
III MIRAGE trial suggest that MR guidance and tracking, had statistically significant improvement in their 2-year
without adaptation, reduced acute grade 2+ genitourinary overall survival rate of 49 % versus 30 %. Overall, no patients
toxicity from 47.1 % to 22.4 % (P = 0.01) and acute GI grade in the high-dose group had grade 3+ adverse events, and
2+ toxicity from 13.7 % to 0 % (P = 0.01) (Kishan et al., three patients in the low-dose group had grade 3+ adverse
2022a). events. The current ongoing Stereotactic MRI-guided
Future studies are likely to include boosting a dominant On-table Adaptive Radiation Therapy (SMART) for Locally
tumor lesion under direct vision, a technique shown to Advanced Pancreatic Cancer (NCT03621644) serves to vali-
improve biochemical control in the FLAME trial (Kerkmeijer date the potential benefits of this approach prospectively. As
et al., 2021), and avoidance of OARs preferentially seen on of November 22, 2021, the study has completed accrual.
MRI. Currently recruiting clinical trials and future directions Similarly, a single institutional retrospective series
for MRIgRT research have recently been published and are reported the use of MRI-guided stereotactic ablative radio-
likely to include boosts to MRI-visible disease and more pro- therapy (SABR) for 17 patients with locally advanced chol-
found hypofractionation, perhaps extending down to just two angiocarcinoma. Two-year overall survival and local control
fractions (Westley et al., 2022). The clinical utility of were 46.1 % and 73.3 %, respectively (Luterstein et al.,
MRIgRT will rely on identifying boost indications that are 2020). One patient experienced grade 3 duodenal ulceration
hard to deliver on other platforms, such as boosting locally and a second patient experienced late grade 2 radiation–asso-
advanced disease. This is currently being tested in the ciated gastritis. Overall, the authors concluded that the treat-
AFFIRM trial (NCT05373316). ment was well tolerated with clinical efficacy that was
encouraging, warranting additional studies.
Another promising use of MRIgRT is the treatment of pri-
2.6 Hepatobiliary and Pancreatic mary or oligometastatic lesions in the liver using SABR. The
Malignancies advantages of MRI in delineating a multitude of liver lesions
Traditionally, definitive treatment of hepatobiliary malignan- from primary hepatocellular carcinoma to metastatic disease
cies (pancreas, bile duct, liver) with radiotherapy has been from lung, colorectal, breast, and other primary cancers have
challenging due to inability to deliver adequate absorbed been well studied (Brock, 2011; Voroney et al., 2006; Witt
dose to what are considered radioresistant histologies and et al., 2020). Traditionally, using CT-based image guidance,
radiosensitivity of the neighboring organs at risk. MRIgRT these tumors are not easily visualized without utilizing sur-
has the potential to overcome these challenges by offering rogates of motion, including implanted fiducial markers.
superior soft tissue imaging, real-time motion management Implantation of fiducial markers in the liver has limitations,
without the use of surrogate markers, and the ability to cus- including potential toxicity with risks of infection, bleeding,
tomize daily radiotherapy to optimize dose to the tumor and and a delay in initiating therapy. In addition, when treating
reduce dose to the normal tissue. multiple lesions in the liver, implantation of fiducial markers
For locally advanced pancreatic cancer, an initial multi- also becomes a logistic challenge as it requires implantation
center retrospective study of 44 patients treated with MRIgRT in multiple areas in the liver as well as creating artifacts
was reported (Rudra et al., 2019). In this study, there was themselves when delineating the borders of the lesions
heterogeneity in the type of radiotherapy delivered with (Guckenberger et al., 2012). MRIgRT has clear advantages
varying biologically effective dose (BED)1: conventional, in allowing for real-time visualization of liver tumors as well
hypofractionated, and ablative radiotherapy regimens. Based as allowing for real-time tracking of liver tumors during
on institutional and physician guidelines, adaptive radiother- respiratory-mitigated liver-directed SABR. A multi-institu-
apy was instituted on a fraction-to-fraction basis, often for tional study evaluating the early feasibility and clinical out-
those patients receiving at least hypofractionated radiother- come using this approach has been reported. This study of 26
apy. In general, when the treatment plans violated OAR patients with six hepatocellular carcinomas, two cholangio-
carcinoma, and 18 metastatic liver lesions (44 % colorectal
1. ICRU recommends the use of equieffective dose (EQDXα/β) metastasis) with a median follow-up period of 21.2 months
rather than BED (Bentzen et al., 2012); ICRU Report 96 showed that the treatments were well tolerated with
(Sgouros et al., 2021). excellent local control (Rosenberg et al., 2019). Median
 Promising Applications for MRIgRT 23

prescribed dose was 50 Gy delivered in five fractions 2.0 to 2.5 cm margins from the postoperative volume are
although daily dose varied based on proximity to OARs and needed to generate the combined clinical target volume
liver function. At the median follow-up time, the freedom (CTV) and PTV to account for microscopic disease exten-
from local progression was 80.4 %, with 100 % local control sion, daily setup uncertainties, and chest wall motion
for hepatocellular carcinoma, consistent with outcomes in (Meattini et al., 2020). Relative to brachytherapy and intra-
the literature. Overall, no grade 4 or higher gastrointestinal operative approaches, external beam radiotherapy often
toxicities were observed in this study. Similarly, a recent pro- requires a significantly larger PTV margin. MRIgRT is able
spective phase I trial using ablative radiotherapy for liver to reduce the PTV margin by improving the direct visualiza-
tumors using stereotactic MRI guidance was reported. tion of the lumpectomy cavity, minimizing daily patient
Twenty patients (eight primary, 12 secondary) and 25 liver setup errors, and offering beam-on cine-based gated treat-
tumors underwent SBRT using MRI guidance to a median ment delivery. The PTV margins can be reduced to 1.0 cm
dose of 54 Gy in three fractions. The overall 1- and 2-year with MRIgRT, which would be comparable with margins
local control were 94.7 % and 79.6 %, respectively. No acute used during intracavity brachytherapy (Berlangieri et al.,
grade 3 or higher GI toxicities were seen, and one late grade 2022). Section C.4 shows a clinical example of a case treated
3 and separately a late grade 4 toxicity was observed sug- with MRIgRT to a partial breast volume.
gesting favorable efficacy and toxicity profile using this Two recent trials using single-fraction APBI, mostly
approach (van Dams et al., 2022). delivered with MRIgRT, have shown feasibility with early
promising clinical outcomes (Kennedy et al., 2020; Vasmel
et al., 2020). In the first study, an absorbed dose of 20 Gy
2.7 Oligometastases
was given in a single fraction to the lumpectomy cavity in
Concordant with the potential advantages listed above for well-selected patients who were postmenopausal, with estro-
hepatobiliary tumors, oligometastases of the upper abdo- gen receptor positivity and HER2/neu-negative breast cancer
men may also benefit from MRIgRT. Section C.2 outlines with negative margins (>2 mm) (Vasmel et al., 2020). The
a use case for this clinical scenario. Clinical data have PTV margin expansion was 1.0 cm from the surgical resec-
demonstrated feasibility and tolerability of MRIgRT for tion cavity with 5.0 mm truncated from the skin. No grade 3
oligometastases in the upper abdomen, including the or higher acute toxicities were observed after a 10-month
avoidance of negative quality of life impact after treat- median follow-up period. “Excellent-to-good” cosmesis
ment in an older population (Mazzola et al., 2020). before treatment was rated by 97 % of the patients and 100 %
Dosimetric and clinical studies have shown improvement of the physicians prior to SBRT. These ratings remained the
in PTV coverage and a decrease in OAR violations with same 8 weeks and 6 months after MRIgRT APBI. After
MRIgRT (Cuccia et al., 2021b; Regnery et al., 2022). median follow-up period of 12 months, no local, regional, or
With expanding evidence supporting the use of SBRT for distant recurrences were observed.
oligometastases in some scenarios, multiple target treat-
ments are likely to become more common. An example is
given in section C.5.
2.9 Other Clinical Indications
While head and neck radiotherapy may not lend itself to
rapid recontouring and replanning, intermittent adaptation to
2.8 Breast Cancer account for weight loss and/or tumor shrinkage can improve
In breast cancer, a recent strategy to reduce the volume of the therapeutic ratio of treatment, as outlined in the use case
irradiated tissue while shortening the course of therapy has in section C.6. The use of biologically targeted volumes to
been the use of accelerated partial breast irradiation (APBI) improve cure rates and decrease toxicity is a promising ave-
(Hickey et al., 2016). With this technique, well-selected nue for future research. Direct visualization of the target vol-
early stage, low-risk tumors, APBI may reduce the late tox- ume with MRIgRT creates new opportunities for targeted
icities and improve cosmetic outcomes while maintaining dose escalation as well as dose de-escalation in the appropri-
excellent local control (Meattini et al., 2020; Vicini et al., ate clinical settings.
2019). This technique has gained popularity using intraop- Augmented imaging and the ability to adapt the plan also
erative radiotherapy, or intracavitary brachytherapy where open up new avenues for radiotherapy, including some for
high dose per fraction can be directly applied to the lumpec- nonmalignant disease. Radiotherapy ablation of arrythmo-
tomy cavity with a very small margin. However, one of the genic foci in the treatment of refractory ventricular tachycar-
more convenient ways to deliver APBI is with external beam dia is an area of growing interest with several small series
radiotherapy. With this approach, accurate delineation of the now demonstrating feasibility on standard linear accelerators
postoperative breast cavity is important but can often be and now also on MRIgRT platforms (Carbucicchio et al.,
challenging using a CT-based imaging approach. Additional 2021; Lee et al., 2021; Mayinger et al., 2020).
24 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

2.10 Immobilization, Patient Setup, An additional consideration is the longer in-room times
Repositioning, Patient Comfort, (Acharya et al., 2016; Bertelsen et al., 2019; Raaymakers
et al., 2017) which will impact decisions regarding patient
Claustrophobia
preparation. A time analysis has been conducted to summarize
Where centers have access to diagnostic MRI departments, it the time required for different components of the MRIgRT
is recommended that MRIgRT safety rules and guidelines are workflow, revealing that, across 154 patients, the median
in line with existing MRI guidance such as those provided by setup time was 5 min, low resolution scanning 1 min, and
the American College of Radiology (ACR). It must be high-resolution scanning 3 min. While the ART components
ensured that patients and staff undergo MRI safety screening are described in detail below, beam delivery time is also
to identify potential contraindications such as implants. Any affected by the longer delivery times due to gated procedures
implant must be investigated prior to that individual entering (~17 min), lack of more efficient arc therapy, relative low-dose
zone IV (the magnet room) as defined by the ACR. More rate, and some systems using step and shoot IMRT (Güngör
details regarding implants and risks are in section 7. et al., 2021) versus more time-efficient arc therapy. This is
Prior to the first treatment, a full MRI safety screening important to keep in mind as some patients are routinely
assessment must be performed. For all subsequent appoint- required to have a full bladder will need reduced drinking
ments, confirmation of continuing MRI-linac safety status times or fluid volumes to tolerate the longer treatment session
can be established by completing a reduced-length daily time. Feasibility of maintaining treatment position also needs
assessment. An example is given in Appendix B, with more to be considered. Patients have reported music to be helpful
detail on MRI screening given in section 7. and a distraction because of the need to remain on the treat-
The principles of patient positioning in radiotherapy ment table for a long time (Barnes et al., 2021).
remain important for MRI-linacs, but consistency and daily Claustrophobia has not been reported as an issue in the
differences in patient positioning are less important due to majority of patients and early feasibility reports indicate pos-
the ability to adaptively replan each treatment fraction. itive patient acceptability (Acharya et al., 2016; Raaymakers
Particular considerations regarding patient positioning et al., 2017). Sound protection used should adhere to con-
include the bore size (70 cm), required position of the receive ventional MRI guidelines (Kanal et al., 2013; Medicines and
coil, and the need for MRI Safe or MRI Conditional immobi- Healthcare Products Regulatory Agency, 2015).
lization devices. No device should be brought into the MRI- Active patient participation in gating may help to relieve
linac environment unless it is indicated as MRI safe or patient anxiety during treatment. Patient-directed gating has
conditional (refer to section 7) (Kanal et al., 2013; Medicines been shown to be feasible and effective (Kim et al., 2020b)
and Healthcare Products Regulatory Agency, 2015). These and has been clinically used for delivery of single-fraction
requirements may result in patient position being altered MRI-guided lung radiotherapy (Finazzi et al., 2020).
from conventional radiotherapy (Cuccia et al., 2021a). For Conventional IGRT processes and actions may need to be
example, patients receiving radiotherapy for breast cancer reviewed. Where gross anatomy changes are identified, it
may need to be planned and treated using a different type of may be possible to rectify in processes similar to conven-
breast board. If conventional linac plans are required for tional treatment, for example, asking a prostate cancer patient
back-up treatments, as in the case of breakdown, the impact to pass gas if rectal gas is observed on pretreatment imaging.
of this on the workflow needs to be evaluated. Departments However, the impact on patient scheduling may be greater
and/or tumor sites where MRI is routinely used for conven- due to the already increased treatment time and should,
tional treatment planning will likely require fewer changes. therefore, be considered.
 MRI-Linac Interoperability and Existing MRI-Linac Systems 25

3. MRI-Linac Interoperability and Existing

MRI-Linac Systems

The engineering challenges involved in combining an MRI on the magnitude of the magnetic field in regions away from
system with a linear accelerator are significant, and many the imaging volume where the linac components may need to
thought it would be impossible to achieve. However, after sig- be situated, which will determine how difficult it is to resolve
nificant research and development work, various methods of several of the above issues. In contrast, all vendors have cho-
overcoming the major problems were developed, and no one sen to implement a short waveguide design with no bending
solution is necessarily the “best.” As with all engineering magnet. This simplifies many aspects of the design; how-
developments, there is often a compromise required between ever, it constrains the maximum X-ray energy achievable
the various design requirements to come to a functional solu- within certain bunker size restrictions. At present, all MRI-
tion. This has led the researchers/developers to come to solu- linac implementations have used linacs between 6 and 7 MV.
tions that are similar in some ways and very different in others. Given the strong sensitivity of MRI systems to the sur-
The most obvious of these was the initial use of 60Co as a rounding environment, considerations for siting of an MRI-
radiation source (Mutic and Dempsey, 2014), which was a linac are similar to those provided for diagnostic MRI and
solution that minimized the problems that were faced in mak- MRI simulators. Detailed considerations provided in AAPM
ing the linac work in the MRI environment. There are cur- Report 20 (Bronskill et al., 1986) outline key siting elements
rently three commercial vendors of MRI-linacs (Elekta—Unity, such as vibration testing, location near moving metal, vent-
ViewRay—MRIdian, MagnetTx—Aurora-RT). This chapter ing, powering, and cooling with radiation-therapy-related
will discuss and contrast the designs used by each vendor. guidance provided by AAPM TG-284 (Glide-Hurst et al.,
The major problems faced in terms of making the linac 2021b). The MRI-linac vendor planning guides shall also be
work in the MRI environment consist of (1) achieving accel- consulted for proximity to adjacent medical equipment and
eration in a waveguide in close proximity to a magnetic field, manufacturer-specific siting considerations. Safety consider-
(2) achieving radiation delivery from all angles without sig- ations are also needed such as the presence of a discharge
nificant beam perturbation, (3) getting the RF power source pipe for rapid removal of gas in the event of a magnet quench
(klystron or magnetron) to function close to the magnetic (e.g., helium supply rapidly boiling off). Depending on the
field, and (4) getting the various motors (collimator, MLC, MRI-linac design, the presence of the discharge pipe and
gantry, couch) to work in the magnetic field. To reduce the proper venting are required to ensure safety for rapid removal
impact of the magnet on the linac, the linac components are of gas and prevent asphyxiation. Oxygen monitoring should
placed in a low-field region, and/or passive magnetic shield- be also provided in zone IV (Magnet Room).
ing is used to reduce the effect of the magnet on the linac. Another significant design choice that leads to vastly dif-
The major problems faced in terms of making an MRI ferent designs is whether to align the magnetic field to be
scanner work in the linac environment consist of (1) effect of parallel to the radiation field, or perpendicular to it. A cylin-
the RF power source on image noise, (2) effect of motors on drical magnet design can only be implemented with a per-
image noise, (3) effect of gantry rotation on B0 homogeneity, pendicular radiation configuration, whereas a biplanar
(4) effect of moving jaws and MLC on B0 homogeneity, (5) magnet allows either parallel or perpendicular configuration.
effect of irradiating the RF receive coils, (6) influence of Elekta and ViewRay have chosen the cylindrical design,
various magnets (magnetron, klystron, bend magnet, steer- whereas MagnetTx has chosen a biplanar magnet in the par-
ing magnets) used by the linac on B0 homogeneity, and (7) allel configuration.
safety issues related to venting helium from quenching The characteristics of each of the MRI-linac devices are
superconducting magnets through the radiation shielding. shown in Table 3.1. The final design parameters are shown in
Fundamental design decisions have a huge impact on how these tables, and they play a significant role in determining
difficult the above problems are to solve and may introduce how the systems were designed to achieve full functionality.
problems other than those listed above. For example, mag- This table is not comprehensive, and each vendor will be
netic field strengths of between 0.35 and 1.5 T have been continuously releasing new features, so this should only be
chosen. Field strength and magnet design will have an impact considered a snapshot at the time of publication.
26
Table 3.1. Characteristics of each MRI-linac device as of 2022.
Item Elekta—Unity ViewRay—MRIdian MagnetTx—Aurora-RT

Linac characteristics
Radiation source 7-MV FFF 6-MV FFF 6-MV FFF
SAD (cm) 143.5 90 120
TPR20,10a 0.707 0.648 0.632
MLC configuration Single stack Double stack Single stack
Number of leaves 160 138 120
Leaf width @ isocenter (mm) 7.1 8.30 (effective 4.15) 4.75
Leaf speed @ isocenter (mm/s) 86 40 55
Direction of leaf travel IEC Y direction (CC) IEC X direction (L/R) Rotates with collimator IEC X direction (L/R)
Jaws Orthogonal to MLC None Orthogonal to MLC
Collimator rotation None None ±90°
Max field size @ isocenter (IEC X × IEC Y) 57.4 × 22.0 cm2 27.4 × 24.1 cm2 28.5 cm diameter circle
Maximum dose rate @ isocenter for 10 × 10 field (cGy/min) 425 600 600
IMRT delivery technique Step and shoot Step and shoot Sliding window
Step and shoot
Radiates through cryostat Yes No No
Delivery from full 360° No No Yes
EPID available Yes No No
Requires custom planning system Yes Yes No
MRI characteristics
Magnet strength (T) 1.5 0.35 0.5
Magnet type Superconducting Superconducting High temperature superconducting
MRI configuration Closed bore Split closed bore Open biplanar
Quench pipe required Yes Yes No
Bore size (cm) 70 70 60 × 110 (rotates)
Magnetic field orientation Perpendicular Perpendicular Parallel
Acquisition time Typical 3D = 2 min 3D = <3 min, 3D = 3 min
Fast breath-hold = 18 s Fast 17 s
MRI sequences Most standard sequences as available on a diagnostic T1 weighted, T1 weighted,
1.5 T scanner (e.g., T1 weighted, T2 weighted, T2 weighted, T2 weighted,
fat suppression, compressed sense, b-SSFP, DWI, DWI, b-SSFP
FLAIR) b-SSFP
Cine image frame rate (frame/s) 4-6 4-8 n/a
Cine imaging planes Single and orthogonal Single and orthogonal Beams eye view
Motion management Image-based target tracking and automatic beam gating Image-based target tracking and automatic beam gating Image-based tracking or gating not released

Note. FFF = flattening filter free; SAD = source axis distance; TPR = tissue phantom ratio; MLC = multileaf collimator; EPID = electronic portal image device; b-SSFP = balanced steady state free precession; IEC = International
Electrotechnical Commission; DWI = Diffusion Weighted Imaging.
a
TPR for a depth of 20 cm relative to a reference depth of 10 cm for a 10 × 10 field size at the depth of measurement.
 MRI-Linac Interoperability and Existing MRI-Linac Systems 27

Figure 3.1. Representation of the Elekta Unity configuration. From www.elekta.com/products/radiation-therapy/unity/assets/

LLFMRL220406_EK_WP_Unity_A4_f.pdf. MR = magnetic resonance.

3.1 Elekta—Unity Design linac and MRI performance are not hampered by the static
Considerations magnetic fringe field (Overweg et al., 2009). Conversely, the
homogeneity of the static magnetic field is not significantly
The Unity system is based on the early work done at UMC affected by motion of the gantry or multileaf collimator
Utrecht, the Netherlands. Simultaneous irradiation and MR allowing simultaneous MRI, gantry rotation, and radiation
imaging were demonstrated in an experimental setup without delivery (Jackson et al., 2019). Passive shimming is used to
a rotating gantry in 2009 (Raaymakers et al., 2009). In 2017, bring the system within specifications, and active shimming
the first patients, with vertebral bone metastases, were treated is used to compensate for patient-related B0 effects. The
in the Unity system in UMC Utrecht, prior to the CE certifi- active shimming is not gantry angle–dependent. Elekta, as
cation for commercial release in 2018 (Raaymakers et al., with all MRI-linac vendors, has chosen to use a magnetron
2017). In 2018, the first patients, suffering from pelvic lymph driven linac. The magnetron rotates with the linac and is posi-
node oligometastases, were treated utilizing multifractions tioned such that it sits in the low magnetic field region at all
on the commercially available Unity system, again in UMC gantry angles. To achieve the high-field MRI specification, it
Utrecht, the Netherlands (Werensteijn-Honingh et al., 2019). was necessary to irradiate through the multilayers of vacuum
The Unity system characteristics are shown in Fig. 3.1 and insulation, coil carrier, and the helium-filled compartment of
described in detail in Roberts et al. (2021). As with all MRI- the magnet cryostat that will both attenuate and harden the
linac designs, the key is to get the magnetic field at the posi- photon beam. The magnet and gradient coil are optimized for
tion of the accelerator subsystems to be as low as possible. homogeneous transmission by repositioning the supercon-
The Unity system consists of a single energy, 7 MV, standing ducting coils outside of the mid-transversal plane and using a
waveguide linac, mounted in a rotating ring around the mid- split gradient coil (Overweg et al., 2009). The magnet is
transversal plane of a 1.5 T MRI system. The magnet design linked between the two sides of the gap with a superconduct-
has been optimized without sacrificing the B0 homogeneity ing wire in a conduit. This conduit is centered at a gantry
(Raaymakers et al., 2009). Moreover, to minimize the static angle of 13° and irradiation through it is excluded by the sys-
and low-frequency magnetic interference impact, the super- tem due to the significant dose perturbation that it would
conducting coil configuration is tuned to create a low mag- cause. The exclusion zone depends on the target location and
netic field toroid around the magnet in which the most the field size to ensure no part of the beam enters through the
sensitive linac components can be placed. This way both the cryostat pipe. It is also recommended that users avoid
28 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

planning with radiation beams that enter through high-density

components of the treatment couch, although this is not
strictly enforced. If followed, the exclusion angles necessary
to avoid beams entering the high-density components of the
couch also depend on target location and field size. The treat-
ment planning system accounts for this exclusion and for the
slightly heterogeneous beam transmission with gantry angle
by the cryostat and couch transmission. The radiation beam is
always perpendicular to the main magnetic field (B0). Thus,
the Lorentz force on the secondary electrons scattered in the
forward direction will always be maximum in this configura-
tion, which leads to the Electron Return Effect (ERE) and the
Electron Streaming Effect (ESE). As will be described later,
this is similar to the MRIdian configuration. With the Unity
system’s higher magnetic field strength compared with
MRIdian, the radius of curvature for secondary electrons in
low-density material will be smaller, which will cause the
dose perturbation from the ERE to take place at a smaller dis- Figure 3.2. Absorbed dose versus depth distributions for a 10
tance from the beam than for lower field systems, and the × 10 cm field size for Magnetic Resonance Imaging (MRI)–guided
Radiation Therapy systems. Elekta data provided by University
dose difference will be more intense with the higher field of Iowa Health Care, ViewRay data provided by Henry Ford
strength. The source axis distance (SAD) achievable in their Hospital, and MagnetTx data provided by Cross Cancer Institute.
design is 143.5 cm. Due to the inverse square law, this SSD = source surface distance.
extended SAD along with the attenuation from the cryostat
limits on the dose rate that can be achieved with this design. The Unity system bore size is 70 cm in diameter. The
The increased SAD and increased beam energy combined couch top does not have lateral or vertical translation capa-
mean the absorbed dose versus depth distribution from the bilities, which prevents the patient from being moved into
Unity system appears more penetrating than the others. Note treatment position if pretreatment imaging detects an offset
that the absorbed dose versus depth distribution has both an from the treatment plan. Instead of moving the patient into
inverse square component and an attenuation component that position, the treatment plan is adapted to the new patient
exaggerates the difference in energy. A comparison of the position prior to treatment delivery. This process is
absorbed dose versus depth distribution for all the MRI-linacs described in detail in section 6. With the wide field size
is shown in Fig. 3.2, where they are measured with a Source possible on the Unity system, peripheral tumors can be
Surface Distance (SSD) equal to its SAD-10 cm for Elekta treated while the isocenter is centrally located by utilizing
and MagnetTx and SAD-5 cm for ViewRay. This shows the asymmetric fields. An evaluation of whole breast irradia-
Unity system’s beam is more penetrating than the MRIdian tion and partial breast irradiation has been done (van Heijst
and Aurora-RT systems. As can be seen in Table 3.1, the et al., 2013) showing the magnetic field will cause a signifi-
TPR20,10 (Tissue Phantom Ratio for a depth of 20 cm relative cant increase in skin dose for whole breast irradiation. For
to a reference depth of 10 cm) is higher than for the other two partial breast irradiation, the impact of the magnetic field is
machines also indicating the higher beam energy. As the relatively minor. This effect will be present for both the
Unity system’s radiation beam is perpendicular to the mag- Unity and MRIdian systems.
netic field, the surface dose is very low, as the contaminant Utilizing a 1.5 T MRI system gives the Unity system a
electrons get swept out of the beam; however, these electrons higher signal compared with the other vendors. Image
do contribute to the ESE. The Unity system utilizes the colli- quality in MR is strongly dependent on signal-to-noise
mation system similar to that of an Elekta Agility linac; how- ratio (SNR) and contrast-to-noise ratio (CNR). Minimizing
ever, with the extended SAD, the conventional 40 cm the noise created in the MRI-linac environment is a key to
maximum field width becomes a maximum of 57.4 cm in the achieving good SNR and CNR. Minimizing the noise is
IEC X (left / right) direction, and their MLC leaf width much more difficult in a radiotherapy environment than it
becomes 7.1 mm at this SAD. The field length in the cranial is in a diagnostic imaging environment due to all the linac
caudal direction is limited to 22 cm by the gap between the components that can potentially cause RF noise. Using a
superconducting coils. The MLC leaf motion is in the cranial Faraday cage is a standard method of separating the elec-
caudal direction with relatively high speed with the future trically noisy components from the MR environment. The
intention of MLC tracking. The Unity system also includes an Unity system’s Faraday cage layout is unique in that the
integrated beam stop that reduces the required shielding in the two sides of the Faraday cage are connected via the metal,
vault. inner wall of the cryostat. This way the gantry, with linac
 MRI-Linac Interoperability and Existing MRI-Linac Systems 29

Figure 3.3. (A) Schematic drawing of the MRIdian system depicting the main hardware components: superconducting double donut
magnet, circular radiation gantry and patient couch and (B) schematic drawing of the radiation gantry with linear accelerator components
and MLC. MLC = multileaf collimator.

and all its peripherals, is outside of the Faraday cage pro- Lorentz force on the secondary electrons will always be
viding the required RF separation to enable low signal maximum in this configuration, which leads to the ERE and
imaging such as diffusion-weighted imaging. ESE. However, since the radius of curvature for the second-
ary electrons will be larger with the lower field strength as
compared with the Unity system, the ERE will be less pro-
3.2 ViewRay MRIdian Design nounced than it is for the Unity. As with the Unity system,
Considerations the perpendicular configuration means the skin dose will be
very low as the contaminant electrons are swept out of the
The ViewRay MRIdian system is based on early work from beam. This does, however, result in the ESE that will be
the University of Florida with the first clinical implementa- similar between the MRIdian and Unity systems (Malkov
tion described in the literature in 2014 (Mutic and Dempsey, et al., 2019). Like the Unity system, the MRIdian linac
2014). The MRIdian system was the first commercial MRI- rotates around the stationary magnet, and the magnetron
guided external beam radiotherapy device coupling three rotates on the same gantry. However, in the case of the
60
Co heads mounted 120° apart on a ring gantry and later MRIdian system, the magnet is split such that the photon
converted into a linac-based delivery system. beam does not pass through any part of the magnet as it
The ViewRay MRIdian system uses the lowest magnetic traverses the gap between the target and the patient. The
field strength (0.35 T) of all the currently available MRI- MRIdian linac system must rotate inside the support struc-
linac vendors. The magnet uses a double donut horizontal ture that holds the two cryostats apart. This leads to the
design with a vertical gap. To isolate the linac and MRI MRIdian having an SAD of 90 cm, which is much smaller
components from each other, ViewRay’s MRI-linac design than for the systems of the other vendors. The 90 cm SAD
incorporates shielded components mounted around the gan- does influence the absorbed dose versus depth distribution
try that include concentric ferromagnetic cylinders that as can be seen in Fig. 3.2 where the beam appears less pen-
house the linac and magnetron (Klüter, 2019) as well as all etrating than that of the other vendors. However, the
other components in the RF chain. A schematic representa- TPR20,10 value shown in Table 3.1 is quite similar between
tion of the system along with a drawing showing the radia- the Aurora-RT and the MRIdian, indicating that the ener-
tion producing components on the ring gantry can be seen gies are very similar, but lower than for the Unity system.
in Fig. 3.3. The RF shielding consisting of RF-absorbing The gap between the two magnet sections does restrict the
carbon fiber and RF-reflecting copper layers is also present maximum field length in the craniocaudal direction, which
to reduce the MRI system from RF influences. The radia- is 24.1 cm on the MRIdian system. ViewRay has imple-
tion beam is always perpendicular to the B0. Thus, the mented a double-stack MLC configuration in the MRIdian
30 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

where the two banks of MLCs are offset by half a leaf beam to be aligned with the main B0 field. The magnet and
width, thereby yielding extremely low interleaf leakage and linac rotate in unison, keeping the linac aligned to B0 at all
an effective leaf width that is essentially half of the full leaf gantry angles. This is illustrated in Fig. 3.4, where the gantry
width. is shown at 330°, and both the magnetic field lines and beam
Both the MRIdian and Unity systems have a 70 cm diam- path are rotated by this amount. This means that the Lorentz
eter bore. In addition to the longitudinal direction, the force on the secondary electrons will be much less than with
MRIdian couch translates in both the lateral and vertical the other two systems, and the dose perturbation caused by
directions, so this allows the patient to be moved into the cor- the main magnetic field is minimal. Thus, the Aurora-RT
rect position if any misalignment is found during pretreat- system does not exhibit any significant ERE (Kirkby et al.,
ment imaging providing that collisions between the patient 2008). MagnetTx utilizes a 6-MV linac, and the beam is
and bore are avoided. Generally speaking, patients are posi- aligned with an opening in the magnet pole that allows an
tioned with their treatment site as close to the magnet isocen- unobstructed beam path from the target to the patient. The
ter as possible to improve treatment efficiency and reduce Aurora-RT system utilizes a custom pole plate to provide
geometric distortions. passive shielding to bring the magnetic fringe field to near
The MRIdian system has the lowest field strength of all zero around the linac. The SAD for the Aurora-RT system is
vendors at 0.35 T; thus, the SNR will be lower than the higher 120 cm, which is between that of the MRIdian and Unity
field strength counterparts. To achieve clinically acceptable systems. The maximum absorbed dose rate provided by the
SNR and CNR, surface phased-array receive coils (12-ele- Aurora-RT system is 600 cGy/min at isocenter, which is the
ment torso and 10-element head and neck) are placed on the same as the MRIdian system. The opening in the magnet lim-
anterior and posterior of the patient. Furthermore, efforts are its the maximum field size that can be achieved with the
made to reduce RF interference in the vault by using a stan- Aurora-RT system to a 28.5 cm diameter circle. Unlike the
dard RF cage surrounding the entire room and with noise- other vendors, MagnetTx has chosen to leave the full circular
generating equipment outside of the RF cage. However, the field available for use rather than limit it to a maximum
linac, magnetron, and associated electronics are gantry square or rectangular field. The MLC leaf width as projected
mounted and need to be shielded. The MRIdian system to isocenter is 4.75 mm, which is smaller than the other man-
incorporates an RF cage around the linac and another around ufacturers although slightly larger than the effective width of
the magnetron individually. Similar to the design of the Unity the MRIdian system. The absorbed dose versus depth distri-
MRI-linac, the rotating gantry for the linac causes metal bution for the Aurora-RT system is shown in Fig. 3.2. Due to
components to be rotating relative to the static MR magnetic the extended SAD compared with the MRIdian system and
field, which causes B0 inhomogeneities. The MRIdian sys- shorter SAD compared with the Unity system, the Aurora-RT
tem also incorporates gantry-angle-specific active shimming system sits between the two in terms of penetration. Fig. 3.2
to compensate for these distortions that have been shown to demonstrates that the surface dose on the Aurora-RT system
influence the location of the imaging isocenter (Lewis et al., is higher than that of the Unity and MRIdian systems. This is
2021b) and diffusion-weighted imaging (Lewis et al., due to the contaminant electrons being captured and follow-
2021a). Full gantry rotation is limited in that it cannot rotate ing the magnetic field lines down to the patient in the
through the region between 30° and 33°. Aurora-RT system, whereas in the Unity and MRIdian sys-
tems the magnetic field is perpendicular and sweeps the con-
3.3 MagnetTx Aurora-RT Design taminant electrons away lowering the surface dose. While
this is not ESE by its typical definition, the increased skin
Considerations
dose within the field is caused by electrons streaming to the
The MagnetTx Aurora-RT MRI-linac is based on the work surface in much the same manner as in ESE. However, in the
from the Cross Cancer Institute and University of Alberta in Aurora-RT the increase happens within the field and is
Edmonton, Canada. They were the first to show proof of caused by contaminant electrons from the linac as opposed to
principle for linac-based MRIgRT by combining a 6-MV the Unity and MRIdian where the increased dose is outside
linac with a biplanar 0.2 T MR imaging system in 2008 the field and comes from Compton electrons set in motion in
(Fallone et al., 2009). the patient (Oborn et al., 2012). The Aurora-RT system
The Aurora-RT system utilizes a significantly different includes an integrated beam stop that minimizes the vault
design compared with the Unity and MRIdian systems. A radiation shielding requirements.
representation of the Aurora-RT configuration is shown in The biplanar magnet design means that the bore of the
Fig. 3.4. Instead of a cylindrical design, the Aurora-RT sys- Aurora-RT system is an elliptical shape with minor and
tem utilizes a rotating biplanar magnet design. The magnetic major axes of 60 × 110 cm. This allows for significant couch
field strength of 0.5 T lies between that of the Unity and the translations where the couch can move ±23 cm laterally and
MRIdian systems. The biplanar design allows the radiation ±35 cm vertically. The orientation of the 60 cm × 110 cm
 MRI-Linac Interoperability and Existing MRI-Linac Systems 31

Figure 3.4. Representation of the MagnetTx Aurora-RT configuration showing the biplanar superconducting magnet mounted on the
same gantry as the linear accelerator such that they rotate together keeping the magnetic field parallel to the beam path.

opening rotates with the gantry rotation; thus, the translation account for this as well as any deviations caused by
along the 110 cm axis is gantry angle–dependent. This allows changes in the magnetic environment at different gantry
the patient to be translated to bring the tumor to the central angles. To minimize the environmental effects with gantry
axis of the beam at any gantry angle even for peripheral rotation, MagnetTx also specifies that silicon steel be
tumor locations. placed in the walls, floor, and ceiling as magnetic shield-
The Aurora-RT system utilizes a 0.5 T high temperature ing. The magnetic shielding sits between the RF cage and
superconducting magnet. The operating temperature must the concrete radiation shielding.
be less than 18 K to maintain superconductivity. This
allows the system to be cooled through conductive cool- 3.4 Australian MRI-Linac Design
ing, and no liquid cryogens are used in the magnet.
Therefore, unlike the other systems, the Aurora-RT MRI-
Considerations
linac does not require a quench pipe that should simplify The fourth MRI-linac design built to date is that from the
vault construction requirements. At 0.5 T, the Aurora-RT Australian MRI-linac Program (Keall et al., 2014) that
field strength sits below the Unity and above the MRIdian designed and built a 1.0 T split bore inline system. The
systems. The Aurora-RT system utilizes an RF cage that active shielding of the magnet was designed to have a low
surrounds the entire machine, and electromagnetic noise– field where the linac is placed, to reduce the impact of the
generating electrical devices are moved as much as possi- MRI on the linac and vice versa. The beam generation and
ble outside of the RF cage to reduce the amount of noise. field shaping are from a 6-MV industrial linac and 120-
With the rotating magnet design, the linac and magnet are leaf MLC placed along B0. One of the unique features of
stationary with respect to each other; therefore, the B0 field the system is the ability to vary the distance between the
is not distorted by the movement of the linac. However, linac and MRI, with eight different locations between 1.82
there is a small influence due to the earth’s magnetic field and 3.27 m allowing studies in different fringe fields and
being at different orientations compared with B0 when the evaluating the impact of the fringe field on the electron
gantry rotates. The Aurora-RT system also includes active contamination within the beam. The magnet has an 82 cm
shimming magnets that are gantry angle–dependent to diameter bore with a 50-cm gap between the two magnet
32 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

halves. The 50-cm gap is less than that of most modern

MRI and CT scanners, and is a tradeoff between patient
inclusion and comfort, where a larger gap is desired, with
image quality, where a smaller gap is desired. The maxi-
mum field of view is 30 cm. The focus of the system has
been for research, with published studies including pre-
clinical nanoparticle theranostic imaging (Byrne et al.,
2020), real-time image guidance and simultaneous multi-
target beam targeting (Liu et al., 2020), radiation detector
development (Alnaghy et al., 2020), and RF coil engi-
neering. An automated horizontal patient rotation system
enables multiple beam angles, superior-inferior and verti-
cal translation of the patient. The Australian MRI-linac in
its current form is unlikely to be used for broader clinical
implementation, and therefore the system is included in
Figure 3.5. Representation of the Australian magnetic
this report for completeness, however, will be given less resonance imaging-linac configuration showing the split-bore
emphasis than other systems. A representation of the MRI- superconducting magnet with the linear accelerator beam path
linac is shown in Fig. 3.5. aligned with the magnetic field. Courtesy of Dr Bradley Oborn.
 MRIgRT Treatment Planning 33

4. MRIgRT Treatment Planning

4.1 Treatment Planning System B = 0.2 T while considerable absorbed-dose increases were
Commissioning observed for all the other field strengths. The orientation of
the magnetic field can be inline (e.g., the MRI scanner and
The AAPM TG-53 report outlines comprehensive QA for
linac axes of symmetry coincide) or perpendicular (e.g.,
initial Treatment Planning System (TPS) implementation,
treatment beam perpendicular to the main magnetic field)
including key considerations such as machine descriptions,
(Constantin et al., 2011). The orientation has a dramatic
field and MLC shapes, verification of photon calculations,
effect on the ERE, with each having different advantages/
and input/output configurations (Fraass et al., 1998). For
disadvantages, including dependencies on field strength,
MRIgRT, consideration must also be given to the presence of
radius of gyration, and tissue type (Keall et al., 2014). The
the magnetic field and the potential impact on the absorbed-
inline orientation focuses beam-generated electrons toward
dose calculation, particularly at interfaces, as described in
the patient surface in the beam direction. Another consider-
the next section.
ation is the surface angle of entry (both exit and entry sides)
with effects mitigated if 1 cm exit bolus or higher magnetic
4.2 Absorbed-Dose Calculation fields combined with large positive surface angles are used,
respectively (Oborn et al., 2010).
For MRI-guided photon therapy, the photon beam is not As a result of the differences in geometry and magnetic
affected by the magnetic field although the secondary elec- field, correction-based absorbed-dose calculation algorithms
trons that are responsible for absorbed-dose deposition are are not applicable. For the perpendicular orientation,
affected. The Elekta Unity system has an SAD of 143.5 cm MRIgRT requires dedicated absorbed-dose calculation
with a maximum field size of 57.4 cm by 22.0 cm. The approaches as well as appropriate beam modeling that takes
ViewRay MRIdian system, in contrast, has an SAD of 90 cm the magnetic field into account. Monte Carlo–based
with a maximum field size of 24.1 cm by 27.4 cm at the iso- absorbed-dose calculation algorithms are well suited to
center. Typical effects include asymmetric absorbed-dose MRIgRT treatment planning and optimization (Ahmad et al.,
deposition perpendicular to the magnetic field, a small 2016; Paudel et al., 2016; Raaymakers et al., 2017).
increase in the beam penumbra for large field sizes, and a
shift of the absorbed dose versus depth distribution. The
most noteworthy effect that requires modeling is the ERE 4.3 Treatment Planning Strategies
(Raaijmakers et al., 2008) that occurs if the beam is perpen- As daily online adaptation takes longer than conventional
dicular to the magnetic field as is the case for the MRIdian radiation therapy and the MRI bore geometry may be confin-
and Unity units. The ERE occurs when an electron exits a ing, patient comfort and compliance are key factors in patient
dense material into a less dense material such as air with suf- positioning. As a result, some groups have integrated per-
ficient remaining range to turn and reenter the previous forming simulation with the patient arms positioned down
material, perturbing doses up to 40 % (Oborn et al., 2009; and eliminating beam geometries that traverse the arms. For
Raaymakers et al., 2008; Rubinstein et al., 2015). The ERE an oligometastatic liver disease cohort of 15 patients, compa-
also affects air-filled dosimeters (Malkov and Rogers, 2018; rable GTV coverage was found between planning with an
Meijsing et al., 2009; O’Brien et al., 2016; 2018; Reynolds MRI-linac (15 beam IMRT) with the arms positioned down
et al., 2017). Monte Carlo–based dose simulations (Oborn and treatment beams avoiding the arms, and dual-arc volu-
et al., 2012; 2016; Park et al., 2016; Yang et al., 2015) are metric modulated arc therapy using a conventional treatment
primarily used to evaluate the magnetic field effect on the plan with the arms up (van den Wollenberg et al., 2019).
absorbed-dose distribution, revealing that magnetic field– However, absorbed-dose conformality was reduced with
induced absorbed-dose effects depend on the field orienta- arms down and OAR doses were slightly compromised.
tion, field strengths, and beam energy. Raaijmakers et al. Another factor when considering an arms-down scenario is
(Raaijmakers et al., 2008) compared the ERE dose changes the need to ensure reproducibility of arm position for treat-
with different field strengths (B = 0.2 T, 0.75 T, 1.5 T, and ment and to assess beam arrangement clearance on a daily
3 T) and determined that the ERE had minimal impact for basis.
34 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

At present, MRIgRT is limited to static beam angles for Two-dimensional (2D) DIR has also been employed to
treatment planning with all three available systems offering determine correlations of voxel-by-voxel shifts between cine
step-and-shoot IMRT. Ongoing research has been conducted images and a reference key frame for gating the linac beam.
to evaluate the potential of arc therapy for the Elekta Unity While DIR is prominent in MRI-linac deliveries and online
system through modifications of the control system software ART, rigid registration still has considerable value for estab-
accomplished via accepting a dynamic stream of control lishing couch shifts in IGRT and for directly mapping rigid
points when sending the treatment plan to the linac, a custom organs between fractions. Comprehensive reviews of DIR
image processing algorithm, and real-time delivery. Overall, are available in the literature (Glide-Hurst et al., 2021a; Shi
comparable plan quality to conventional radiation therapy et al., 2021).
plans and improved delivery efficiency were realized
(Kontaxis et al., 2021). On the ViewRay MRI-linac, eddy
4.5 MRI-Only Planning
currents and RF interference between the moving gantry and
MLCs degrade MR image quality, thereby prohibiting arc Using MRI as the primary image for planning and treatment
therapy; however, this is an active area of inquiry. delivery is advantageous for initial patient localization (i.e.,
Much like conventional linacs, couch attenuation needs to single imaging modality registration) and is required for
be accounted for in the TPS, with some end users avoiding adaptive radiation therapy in many currently available work-
beams with long path lengths at the outer edges of the couch. flows. However, no fundamental relationship exists between
Attenuation through the RF surface coils can be modeled MRI intensity values and the electron density information
within the TPS (often a bulk attenuation correction) and the required by treatment planning systems to accurately account
presence of immobilization devices should also be accounted for tissue heterogeneities for accurate absorbed-dose calcula-
for in the absorbed-dose calculation as per AAPM TG-176 tion. Thus, a necessary component for MRI-only planning is
(Olch et al., 2014). Overall, beam angle selection should be the generation of a synthetic CT from MRI data. Table 4.1
made with considerations for delivery efficiency and plan outlines key considerations for MRI-only planning, typical
quality. uncertainties, and their associated clinical consequences.
The simplest way to generate synthetic CTs is to assume a
homogeneous patient and assign the body to a water unit
4.4 Deformable Image Registration density, although this simplification has been shown to intro-
Deformable image registration (DIR) may be used during duce ~2.5 % dosimetric error in prostate radiotherapy (Chen
many key aspects of MRIgRT, including contour propaga- et al., 2004). This simplification cannot be generalized to all
tion employed to account for interfraction anatomical varia- sites as errors of over 10 % would be made for some thoracic
tions that arise during the treatment course and to perform an cancer treatments. Assigning other density overrides such as
electron density mapping between the planning image and air and bone in segmented regions has been shown to improve
the daily patient-positioning images. Deformation vector absorbed-dose uncertainty to <1.5 %.
fields (DVFs) are obtained via the DIR process and consist In some clinics, a hybrid approach is used where a syn-
of a voxel-by-voxel three-dimensional (3D) transformation thetic CT is generated for the initial treatment plan based off
matrix that are applied for contour propagation, image warp- of the MR-SIM such as Philips Magnetic Resonance for
ing, and dose accumulation in an online ART workflow. Calculating Attenuation (MR-CAT) (based on a dual echo
However, DVFs may contain errors that arise from aspects 3D mDixon fast field echo sequence) (Köhler et al., 2015;
such as differences in image FOVs, dissimilar data sets (e.g., Tyagi et al., 2017) while the corresponding online MRgART
highly variable anatomy or resolution/modality differences), workflows rely on stratified synthetic CTs using three major
and poor image quality, with the uncertainty propagating electron density (ED) classes: soft tissue (ED = 1), femoral
throughout downstream tasks. While qualitative review of bones (ED = the average ED from a pretreatment CT), and
the image or known landmarks is often valuable to under- (external) air (ED = 0) while also taking into account couch
stand regions with a high degree of uncertainty, several quan- attenuation and transient gaseous regions (Godoy Scripes
titative metrics of DIR performance (e.g., target registration et al., 2020; Intven et al., 2021). In addition to accurate ED
error, mean distance to agreement, Dice similarity coeffi- assignments, the presence of air cavities may influence hot
cient, the Jacobian matrix, and consistency) have been spots at air cavity interfaces as has been observed with ~1 cm
described in detail in AAPM TG-132 with expected toler- air cavity due to the ERE in a 1.5 T MRI-linac setting (Godoy
ances (Brock et al., 2017). Practically speaking, end users Scripes et al., 2020). Yet, these population-based approaches
can make manual or semi-automatic corrections to erroneous may not account for air such as rectal gas in the synthetic CT
contours or improperly assigned electron densities, which data set that has been shown to cause absorbed-dose calcula-
should be incorporated into the initial and adaptive treatment tion discrepancies (Kim et al., 2015) thus requiring manual
plan reviews before treatment delivery. overrides. Another commonly employed strategy often
 Table 4.1. Key considerations for MRI-only planning, typical uncertainties, and their associated clinical consequences.

Class Element Suggested minimum requirement Potential clinical impact

Acquisition efficiency If breath-held, generation in <20 s Impedes workflow, decreased patient compliance
Anatomical representation Deemed adequate for treatment planning per disease site Overestimation/underestimation of anatomical
and use volumes
Discrepancies in treatment planning
Clinical Consistent physiological state as reference Motion managed within AAPM TG-76 recommendations Overestimation/underestimation of target/OAR doses
workflow data set (breath-hold/internal filling) (<5 mm) (Keall et al., 2006) Incorrect state of internal anatomy for treatment
planning
Localization using clinically applicable ART Data integrity verified via AAPM TG-53 (Fraass et al., 1998) Localization uncertainties
workflows and imaging modalities System-specific concordance of external laser system/ Input/output discrepancies
landmarks: Systematic offsets introduced
Preferred: 1 mm
Acceptable: 2 mm
Geometric integrity ≤1 mm (within 10 cm radial distance of isocenter) Inaccurate localization of organs
≤2 mm (>10 cm radial distance away from isocenter) Inaccuracies in absorbed-dose calculation
(Baldwin et al., 2007)
Low-contrast resolution Per AAPM TG recommendations for the ART planning Limited boundary detection that may adversely impact
modality accurate delineation on MRI
Image Consistent physiological state as reference Motion managed if > 5 mm per AAPM TG-76 Overestimation/underestimation of target/OAR
quality data set (breath-hold/internal filling) recommendations (Keall et al., 2006) absorbed doses
Incorrect state of internal anatomy for treatment
planning
Field of view Contains all relevant anatomy and full integrity skin contour Inaccurate absorbed-dose calculation for missing
anatomy
Lack of one-to-one correspondence may lead to
erroneous deformable image registration
Artifacts Initial and ART treatment planning images shall be free of May obscure relevant anatomy
artifacts in the clinically usable field-of-view. Delineation accuracy adversely impacted
Absorbed-dose calculation may be adversely impacted
Dosimetric Hounsfield unit/electron density accuracy CT number accuracy within 10% 20% variation in Hounsfield unit value may result in a
impact systematic dose error of 1.5% (Zurl et al., 2014)
Dosimetric accuracy ±5% of intended absorbed dose (Fraass et al., 1998) Inaccurate absorbed-dose evaluation
Note. ART = adaptive radiation therapy.

35
36 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

applied for ED map generation for dose calculation relies 4 mm lateral/dorsal and 6 mm craniocaudal/ventral) and the
upon DIR that is applied between daily MRI data sets and the elective nodal regions (8 mm reduced to 4 mm lateral/ventro-
original MR-SIM data set to generate a daily DVFs. These dorsal and 6 mm craniocaudally) (Intven et al., 2021). In a
DVFs are then applied to the original CT-SIM data set for recent randomized phase 3 single-center trial (MIRAGE),
daily dose calculation and any inconsistencies in tissue/air SBRT for localized prostate cancer, patients were randomly
are manually corrected to reflect the anatomy of the day. As assigned to either CT guidance or MRI guidance with plan-
this process is labor-intensive and relies upon DIR perfor- ning margins of 4 mm (CT-arm) and 2 mm (MRI-arm)
mance, efforts have recently been made to develop deep around the prostate and proximal seminal vesicles, with the
learning solutions using MRIgRT data, such as using genera- rationale of margin reduction secondary to the soft tissue
tive adversarial networks (GANs) and conditional GANs contrast and online monitoring available in the MRIgRT
with paired MRI/CT data in lung, pelvic, and abdominal can- arm. Interim analysis has shown significant reductions in
cers with promising results (Cusumano et al., 2020; acute grade ≥2 GI, and GU toxicities were significantly
Lenkowicz et al., 2022) with more accurate synthetic CTs reduced in the MRIgRT arm (Kishan et al., 2022b).
than conventional neural networks (Emami et al., 2018) In addition, the low-field MRI-linac and its predecessor
60
although future implementations within clinical workflows Co MRI-linac enabled near real-time gating based on tumor
are needed. or surrogate tracking in cine images acquired in the sagittal
plane. Fifteen patients (87 treatment fractions) with lung,
4.6 When can we Safely Reduce the adrenal, and pancreas tumors were evaluated using gated
MRIgRT delivery and a visual feedback system, and the
PTV Margin? investigators found that a 3 mm GTV to PTV margin could
It has been well established that the required margin size to account for almost all of the uncertainties in the study (van
maintain overall plan quality depends on the modality of Sörnsen de Koste et al., 2018). Of note is that to maintain
image guidance and frequency of use, with increased use adequate target coverage, overlap criteria are used between a
reducing beam-target alignment errors (Kupelian et al., predefined boundary and the tracked target at the expense of
2008; Zeidan et al., 2007). Through the process of online reduced treatment duty cycle (e.g., time the tumor spends
MRI-guided ART, interfraction setup variations can be within the prescribed boundary). In initial reports of >300
reduced through recontouring and replanning strategies. patients treated with gated MRIgRT, PTV to GTV margins
Further reduction in target margins may be justified through were reduced from 5 mm down to a 3 mm predefined bound-
the improved visualization and potential for real-time motion ary as the GTV itself is used as the gating target (Fischer-
management with continuous feedback afforded by MRI Valuck et al., 2017).
guidance as compared with X-ray based on-board imaging, When considering any change in margin, it is important to
leading to improved confidence in soft tissue targeting (Noel understand the uncertainties are included in the CTV and
et al., 2015). One such example has been observed in a pro- PTV. ICRU Report 83 (2010) provides guidance. Also, there
spective study using a 1.5 T MRI-linac in 43 rectal cancer are variations between MRI- and CT-defined tumor and tar-
patients. After an interim analysis, PTV margins were get volumes (Pathmanathan et al., 2019) that should be
reduced for both the mesorectum (1 cm isotropic reduced to understood.
 Imaging with an MRI-Linac 37

5. Imaging with an MRI-Linac

Table 5.1. Categories of images acquired at different time points during magnetic resonance imaging-guided radiation therapy, with
their purpose in the workflow.

Image type Element (Fig. 1.1) Primary purpose Secondary purpose

Planning images 1 Treatment simulation and Used as a comparison image for subsequent
subsequent treatment planning treatment setup and adaptation
Setup/adaptation 4 Capture the anatomy of the Response monitoring (based on anatomical
day to be used for position images)
correction or plan adaptation Capture and initialize motion model of the
day in case of motion management
Verification 8 Final check of the target Used as a comparison image for anatomic
positioning, prior to starting changes and dose accumulation
irradiation
Motion management 9 Motion management Dose accumulation
Response monitoring 4–9 Monitoring treatment response Outcome prediction

During the MRI-guided radiotherapy chain as represented in

Fig. 1.1, the acquired MR images serve different purposes.
Table 5.1 outlines five categories of images that are acquired
at different time points during MRIgRT, with their typical
purposes in the workflow.

5.1 Image Quality for MRI Guidance

The starting point of online adaptation relies on accurate detec-
tion and assessment of anatomic changes which require higher
image quality than general image guidance for patient position-
ing. The image quality should also be sufficient to allow pre-
cise delineation of the target and organs for potential replanning.
In addition, fast image acquisition must be maintained to mini-
mize the treatment fraction time. Image quality thus relates to
the intended use of the images acquired during treatment on an Figure 5.1. Three-dimensional T2-weighted magnetic resonance
MRIgRT system. For MRIgRT, diagnostic quality is not usu- imaging of a patient with prostate cancer, used for adaptation on
ally necessary, as the diagnosis is already known at the onset of the Unity system (1.2 mm × 1.2 mm × 2.0 mm voxel size, 40 cm
× 45 cm × 25 cm field of view, 118 s acquisition time).
the treatment. The extent of the disease and the relation to nor-
mal structure is often known from CT, MRI, and positron emis-
sion tomography (PET) images obtained during the diagnostic In the workflows as shown in Fig. 1.1, we can distinguish
or pretreatment phase (Fig. 1.1). While it can be relevant to different time points (elements) at which imaging takes
assess changes in the cancer and normal tissue size and shapes place. Table 5.1 lists different categories of images by their
that may prompt a modification of the treatment strategy, the primary purpose (Mutic and Dempsey, 2014). Simulation
use of images acquired during MRIgRT is typically more lim- images can be made during step 1 in Fig. 1.1 on simulators
ited than simulation images. Examples of typical images that including CT simulators, MRI simulators, or PET-CT/PET-
are used for adaptation and verification for the treatment of MRI simulators where patients are imaged in the treatment
prostate cancer on the Unity (Fig. 5.1), MRIdian (Fig. 5.2), and position. Magnetic resonance imaging simulation can also
Aurora-RT (Fig. 5.3) systems are shown. be performed on the MRI-linac if the image quality is
38 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure 5.2. Three-dimensional balanced steady-state free

precession MRI images of a prostate patient acquired on the
MRIdian system (1.5 mm × 1.5 mm × 1.5 mm voxel size, 50 cm
× 45 cm × 43 cm field of view, 172 s acquisition time). Figure 5.3. A three-dimensional radiofrequency spoiled
gradient echo image of a healthy volunteer’s prostate taken on
the Aurora-radiation therapy system (2.08 mm × 1.17 mm ×
4.7 mm voxel size, 40 cm × 30 cm × 30 cm field of view, 154 s
sufficient. Here, images are acquired as needed for radiation
acquisition time).
treatment planning with the patient in the treatment position.
Images can be acquired at the desired physiological position
(e.g., breath-hold or bladder filling) (Glide-Hurst et al., and containing sufficient extent for dose fall off. If the online
2021b). If synthetic CT images can be derived from MRI adaptation process takes a considerable amount of time, it
images acquired at MRI simulation, an MRI-only workflow may be beneficial to acquire verification images to verify
is feasible during MRI-guided radiotherapy as described in that the patient’s anatomy has not changed during the ART
section 4.5. process (elements 5–8 in Fig. 1.1) and the target positioning
The minimum requirement of setup/adaptation images is is still correct. Although in principle the same requirements
that they are suitable for the IGRT alignment purposes for apply, shorter acquisition times with lower image quality
daily treatment. This means that they must have sufficient may be sufficient for verification purposes.
contrast and resolution for the users to discern the known If the target moves during irradiation and the motion man-
target volume and the adjacent normal tissues. If the setup/ agement strategy involves gating or tracking, motion man-
adaptation images are used only for positional correction in a agement images are required that provide the input for that.
manner analogous to IGRT, the quality must be sufficient for At present, motion management largely consists of the acqui-
image registration to the reference data set, which is usually sition of 2D cine-MRI that, depending on platform, can be
an MRI or a CT simulation. To test if the quality of the acquired in the axial, sagittal, or coronal slice direction or in
images is sufficient for this endpoint, it is possible to deter- a combination of subsequent sagittal and coronal images
mine the interobserver consistency of the image registration (Bertholet et al., 2019). This may enable gating of the radia-
or use techniques outlined in AAPM TG-132 for character- tion beam. The contrast is typically a balanced steady-state
izing co-registration uncertainty including qualitative (e.g., free precession (bSSFP) sequence, which is a fast and high-
visual inspection of overlapping images with optional review signal sequence that provides a mix of T1 and T2 contrast.
of corresponding contours) and quantitative (e.g., target reg- Alternatively, a fast T1-weighted gradient echo sequence can
istration error, mean distance to agreement, Dice similarity be used. The clinical systems currently only offer a 2D
coefficient, Jacobian matrix, and consistency) metrics (Brock sequences for motion management due to time constraints.
et al., 2017). However, when motion is infrequent or slow, for example,
If the setup/adaptation images are used for online adapta- for rectum or cervical cancer, the time resolution of the
tion, it may be necessary to recontour target volumes and images may be reduced so that 3D imaging would be feasi-
critical structures. In that case, demands on image quality ble, in particular when combined with acceleration tech-
may be higher to properly localize the target and organs at niques such as compressed sensing (Lustig et al., 2007). The
risk as well as reduce interobserver variability. Generally, benefit of 3D images is that irregular target volumes can be
data sets must be free of image artifacts in the region of inter- tracked more accurately. Moreover, evaluation of the deliv-
est (whether for tumor/OAR localization or dose calculation) ered absorbed dose is more straightforward when 3D images
and cover a large-enough FOV to include necessary land- are available.
marks and anatomy for dose calculation purposes as needed One of the key advantages of MRI-guided radiotherapy is
including accommodating all of the beam angles of interest that at each treatment fraction, the anatomy is imaged so that
 Imaging with an MRI-Linac 39

Figure 5.4. A patient with a rectal-cancer metastasis in the right lateral liver (white arrow) that is hypo-intense on the T2-weighted
coronal slice from a 4D-MRI series (left) and (almost) not visible on the balanced Turbo Field Echo(TFE) coronal slice from a 4D-MRI
series (right).

the day-to-day response of the disease to radiation therapy (2015), showing recommended combinations of MRI
can be observed. Essentially all images that are acquired at sequences that are beneficial for each major disease site. For
any stage during a treatment fraction, that is, in steps 4 to 9 selecting sequences for MRI guidance, a good starting point
of Fig. 1.1, can be used for this purpose. Current MRI-linacs is one of the sequences that would be used in a diagnostic
allow the acquisition of functional and quantitative MRI setting and during MRI simulation for radiotherapy.
sequences, such as diffusion-weighted MRI. Diffusion- Depending on the MRI-linac platform, T2-weighted or
weighted MRI has proven feasible on both the Elekta and bSSFP sequences are quite common. The latter is a fast
ViewRay systems. For the ViewRay system, the apparent sequence with a mix of T1 and T2 contrast, which is often
diffusion coefficient (ADC) values and geometrical fidelity applied for motion management. However, in some cases
varied with changing gantry angle (Lewis et al., 2021a). For T1-weighted, or fat-suppressed sequences are applied. Eccles
the Elekta system, ADC values were not influenced by gan- et al., (2019) presented an early experience from an imaging
try orientation (Kooreman et al., 2020). Typically, these study on healthy volunteers on the Elekta system, suggesting
sequences are not used for online image guidance, but they that for most anatomies, T2-weighted MRI was the preferred
can be used to monitor treatment response. A future perspec- sequence based on visual grading. For thorax, T1-weighted
tive is to use such images to identify the entire tumor or sub- images were preferred (Lustig et al., 2007). To draw defini-
volumes for more precise targeting. tive conclusions however, images of patients are necessary to
While the images made on an MRIgRT system may reveal evaluate visualization of the tumors.
changes related to response, this could prompt scheduling In the current commercially available MRI-linacs typi-
new diagnostic exams such as PET or diagnostic MRI to cally preset sequences are available for setup/adaptation,
investigate the changes in more detail. verification, and motion management. In particular, the Unity
Image quality needs largely depend on the particular dis- system has dedicated sequences with contrasts, resolutions,
ease site and anatomical considerations for the region of and scan durations, tailored to different anatomies and appli-
interest as the visibility of a tumor depends on its character- cations. While standardized sequences support a smooth
istics as well as the specific environment. Therefore, it is workflow, it can be useful to select sequences that are specifi-
quite difficult to make generic statements about optimal cally tailored to each individual case. The examples in Figs.
sequences for image guidance. For MRI simulation, a site- 5.4 and 5.5 illustrate the potential gain of individualization
specific reference guide was published by Paulson et al. and demonstrate the difference between the visualization of
40 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure 5.5. A patient with a metastasis arising from a primary tumor in the appendix, in the medial liver (white arrow) that is less
apparent on the T2-weighted coronal slice from a 4D-MRI series (left) and hyper-intense on the balanced TFE coronal slice from a
4D-MRI series (right).

tumors for the same patient imaged with different sequences. channels in the Unity system (4 anterior/4 posterior) and 12
Both patients had liver metastases and were scanned on the in the MRIdian system (6 anterior/6 posterior for the torso
Unity system with a 4D-MRI scan protocol. This protocol coils and 5/5 for the head and neck coils). The design of the
was developed by the Netherlands Cancer Institute and is not field gradient system and the electronics of the acquisition
available as a standard sequence (van de Lindt et al., 2018). chain further influence the SNR. Compressed sensing and
Contrast agents may be administered to improve tumor parallel imaging–combined techniques are often used to
conspicuity. Recent reports of administering hepatocyte-tar- optimize the acquisition time while preserving the high con-
geted gadolinium contrast (e.g., Eovist/Primovist) have been trast required for adaptation (Gao et al., 2018). Image quality
reported for daily localization of primary or secondary is further influenced by the specifics of the image reconstruc-
tumors of the liver for each MRIgRT fraction with clear ben- tion algorithm. Novel acquisition and reconstruction tech-
efits in tumor discrimination (Gach et al., 2022; Rogowski niques based on deep learning have been published that
et al., 2021). During MRIgRT, the latter should only be improve image quality by increasing SNR (Gassenmaier
applied with caution. While the first evidence is emerging et al., 2021; Terpstra et al., 2021) and may be implemented
that the chemical stability of gadolinium chelates under irra- in MRIgRT systems in the future.
diation seems to be promising (Wang et al., 2021), repeated For any given MRI system, typical image quality trad-
use of contrast agents should be evaluated based on a careful eoffs include scanning time, image resolution, slice thick-
risk/benefit analysis of the patient. ness, and acceptable level of SNR. The ACR Large Phantom
In general, image quality is determined by technical fac- Test Guidance document and scanning of the ACR Large
tors including the strength of the static magnetic field (B0 as Phantom is used for MRI accreditation in many diagnostic
a key factor as the signal increases quadratically with B0 for MRI settings to examine a broad range of instrument param-
a constant acquisition time). However, the noise arising from eters (ACR, 2019). Fig. 5.6 shows images of slice 11 (low
different sources is also influenced by B0. The receive coil contrast object detectability module at the 5.1 % contrast
system is a critical factor, and dedicated coil systems that level) acquired on each of the three commercially available
optimally cover the volume of interest may significantly MRIgRT systems. Nominally, the acquisition is done under
reduce noise. For this reason, multichannel phased array standardized acquisition conditions to test the performance
coils are widely used in the current MRIgRT systems of an MRI platform; however, the low-field MRI of the
although the number of channels is currently limited to 8 MRIdian system deviates substantially from a regular
 Imaging with an MRI-Linac 41

Figure 5.6. Images of slice 11 of the American College of Radiology phantom acquired with a two-dimensional T1-weighted sequence
on the 0.35 T MRIdian (left), 0.5 T Aurora-radiation therapy (center), and 1.5 T Unity (right) systems.

diagnostic system, here 25 signal averages are used to acquire distortions can be minimized to less than 1 mm near the center
these data with the torso surface coils, whereas the Aurora-RT of the image, with tolerances up to 2 mm for the entire FOV.
used 8 signal averages with a 25-cm belt coil, and the ante- Geometrical fidelity is a particular concern for 2D acquisition
rior and posterior surface coils are used with 1 signal average methods, as used in 2D cine motion management. Owing to
on the Unity system. the GNLs, the image plane itself will be warped. Correction
While initial sequence optimization may first be per- algorithms provided by the vendors to date only apply a 2D
formed in healthy volunteers, the sequences for image guid- correction for these 2D acquisitions, leaving the out-of-plane
ance should ideally be finalized by imaging patients with distortion uncorrected. While the distortions may be small for
pathology to ensure that images not only visualize general planes that cross through the isocenter of the MRI system, the
anatomy, but also the cancer itself. For the sequences used out-of-plane distortions can be considerable when imaging
for motion management, it is also preferable that the tumor off-center slices. This should be considered when 2D cine
itself can be visualized, allowing direct tracking/gating of the imaging is used for motion monitoring.
relevant structure. However, while a tumor may be visible on Targets that are subject to breathing motion form a particu-
the (slower) sequence used for setup/adaptation, it some- lar challenge for the acquisition of MR images for setup/
times is not possible to visualize the tumor with either bSSFP adaptation. Breathing motion creates substantial motion arti-
or T1-weighted sequences used for motion management. In facts when acquired under free-breathing conditions. The
that case, it may be an alternative to use a nearby suitable MRIdian MRI-linac relies heavily on the bSSFP sequence
surrogate structure for gating/tracking. that can be acquired in a single breath-hold of 17 to 25 s with
a 3-mm slice thickness and 1.5-mm in-plane resolution
5.2 Technical Requirements of encompassing the patient body contour with clinically accept-
able image quality. When treating during breath-hold condi-
Sequences for MRI Guidance tions, audio or visual feedback is implemented to improve the
Table 5.2 gives an overview of the technical requirements for duty cycle and patient compliance (Tetar et al., 2018). On the
MRI sequences used in MRIgRT. For all images acquired for MRIdian MRI-linac, 2D cine acquisitions at 4 to 8 frames per
MRI guidance, it is critical that they have a high degree of geo- second are used to gate the linac within a predefined gating
metrical fidelity across all axes and within the entire relevant window based on the breath-held condition.
FOV, which AAPM TG-284 defines as distortions ≤2 mm On the Unity MRI-linac, several sequences are available
across 25 cm FOV. However, specific considerations should to deal with breathing motion. Breath-hold sequences can be
be made for each use case (Glide-Hurst et al., 2021b). Without applied, typically in the maximum inhale phase of the breath-
correction, gradient nonlinearity (GNL) is the largest source of ing cycle. For acquisition during free-breathing, navigator-
magnet-specific distortion. However, all MRI systems offer triggered sequences use a 1D acquisition that captures the
2D and 3D gradient corrections. The impact of these correc- position of the diaphragm. This signal can be used to trigger
tions is shown in Fig. 5.7 for Unity. Typically, 3D sequences a 3D sequence, making sure that each repetition is acquired
are used for setup/adaptation and verification imaging. Here a in the same breathing phase. This provides unblurred images
full 3D GNL correction is applied. After such a correction, in the maximum exhale phase. If either breath-hold or
42 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Table 5.2. Technical requirements for MRI sequences in MRIgRT.

Item Measure Criterion Recommendations

Geometrical 3D correction of gradient Residual distortion < 2 mm across Orient treatment site as close to
fidelity nonlinearities implemented by 25 cm FOV magnet isocenter as possible
vendor
2D or 3D correction of gradient Residual distortion < 2 mm in-plane Consider out-of-plane distortions
nonlinearities for 2D cine-MRI across 25 cm FOV when using 2D correction
implemented by vendor
Duration Acquisition time including active - Breath-held < 20 s - If workflow enables verification
shimming - Free-breathing setup/adaptation images, 2 min is often feasible
sequences can typically be acquired - Scan times <30 s may be
in 1–5 min possible for free-breathing,
conventionally fractionated
treatments
Motion during 4D MRI Sufficient to capture the respiratory
acquisition of cycle (at least 4 phases)
setup adaptation Navigator-triggered MRI - Optimized trigger level and Treatment gating must be in the
images trigger delay to acquire at desired same phase as the navigator-
respiratory phase. triggered acquisition
- Acquisition length per trigger event
optimized to ensure prescribed
volume is acquired at specific
respiratory phase at the expense
of increased total acquisition time
(Glide-Hurst et al., 2021b)
3D radial MRI - Minimal artifact In-plane motion artifacts are
Rapid acquisition to accommodate reduced, but cross-plane
breath-hold or triggered from blurring needs to be considered
respiration
Motion 2D cine-MRI Time interval <250 ms between Ideally oriented at a high-contrast
management successive frames is feasible interface at or near the tumor
Note. FOV = field of view.

navigator-triggered sequences are used for setup/adaptation, Images used for dose calculations need to cover the entire
it is critical that the actual treatment is gated in the same patient body contour, for other images a smaller FOV can be
breathing phase. A third solution is to use 3D-radial acquisi- accepted, providing all relevant anatomy is visualized. The
tions to minimize motion artifacts. On the Unity system, this tolerable duration of a sequence depends on its use. Setup/
is available as 3D-Vane. This avoids ghosting artifacts and adaptation images are meant to capture the anatomy of the
results in a visualization of the target in the mean position. day for recontouring. If the anatomy changes, for example,
Treatment during free breathing is now in principle possible. due to peristaltic motion or bladder/rectal status, the prebeam
A drawback is however that the image may be blurred in the images will represent the patient’s anatomy with decreasing
slice direction. As a result, the accuracy of target verification accuracy. Ideally, the total time taken for prebeam imaging,
in the breathing direction may be somewhat compromised. adaptation, and verification prior to starting the irradiation,
Although these solutions are now commercially available, should therefore be short compared with the time scale on
several groups have pursued the generation of 4D MRI which changes in anatomy take place. For most free-breath-
sequences to characterize the full respiratory cycle (Nowee ing applications, prebeam imaging can be done within 5 min
et al., 2021; Stemkens et al., 2018). The solutions that are and high-quality 3D T2-weighted sequences of about 2 min
based on a stack of stars acquisition typically have a are possible. To accommodate breath-held acquisitions,
T1-weighted or balanced contrast while providing acquisitions <20 s are desirable.
T2-weighted contrast is more challenging. One solution is to For motion management imaging, the time resolution is
use standard multi-slice 2D turbo spin echo (TSE) sequences dictated by the time scale at which the anatomy changes and
and sort the slices based on a breathing signal. The 4D images the strategy used for handling the target motion. When track-
can be used to identify the mid-ventilation phase of the ing/gating tumor motion during breathing, cine imaging
breathing cycle or to generate a mid-position image (van de needs to be fast. Currently, a 2D cine image can be acquired
Lindt et al., 2019). This can be used for setup/adaptation. with a reasonable image quality within about 200 ms.
 Imaging with an MRI-Linac 43

are particularly affected by the prosthesis. The magnitude of

the artifact will depend on the material of the prostheses. The
prosthesis shown in Fig. 5.8 has a stainless steel
(Fe-CrNiMnMoNb) stem and stainless steel ring in the cup
as imaged on a 1.5 T MRI-linac. Stainless steel tends to give
larger artifacts than titanium or ceramic prostheses that may
be more commonly implanted. For MRIgRT, prostheses may
present additional challenges. As in non-MRI-guided radio-
therapy, they may impact the selection of beam angles.
However, more importantly, the B0 inhomogeneity caused by
the implants, may distort the images. For adaptation/verifica-
tion images, such distortions may impact the shape and posi-
tion of treatment fields. While these artifacts are not as
prominent at 0.35 T, they are still present.
Other artifacts may be caused by implants outside of the
treatment field but may adversely impact the imaging, such
as the vascular stent shown in Fig. 5.9. These may have less
impact on beam arrangement and planning but may cause
enough signal loss and distortion to affect contouring. Any
geometric distortion to the area to be targeted should be
quantified prior to treatment and this additional uncertainty
factored into the contouring process. For this patient, esti-
mated distortion at the level of the prostate (the target) was
measured to be 1 mm or less. In addition to confounding con-
touring, intensity-based DIR techniques to generate an elec-
Figure 5.7. Large field-of-view landmark phantom with tron density map or contour propagation may also be
embedded fish oil capsules and imaged on a Unity magnetic adversely impacted by the presence of image artifacts.
resonance imaging-linear accelerator with raw data reconstructed
with no vendor-provided gradient nonlinearity corrections, two-
Metallic implants do not necessarily rule out MRIgRT
dimensional (in-plane) corrections, and three-dimensional (3D; in- treatments. Care must be taken, however, to quantify the
and through-plane) distortion corrections applied. Note that the expected geometric distortions caused by the implant to
restoration of the grid pattern in the sagittal plane is improved ensure they are within acceptable limits. B0 field map acqui-
with 3D corrections although some residual uncertainty is still sitions prior to treatment can be used to estimate the magni-
present necessitating a thorough characterization by a qualified tude of distortions during MRIgRT to guide the decision on
medical physicist. eligibility for MRIgRT. The feasibility of this approach was
demonstrated for metal hip implants in patients with prostate
However, for motion occurring over a longer time scale, such cancer (Keesman et al., 2020).
as peristalsis or bladder filling in the pelvis, much longer Importantly, image artifacts may occur during the course
interval times may be acceptable. While flexibility in config- of treatment not necessarily by an implant, but for example
uring cine-MRI in MRIgRT systems is currently limited, this by iron-rich food consumption. As the image quality needs to
would allow a tradeoff between image quality, number of be maintained over the treatment course, care needs to be
slices acquired, and time. As acceleration techniques are fur- taken to prevent image artifacts due to intratreatment
ther improving, fast 3D acquisitions may be suitable for this changes. Fig. 5.10 demonstrates an example of significant
purpose as well. metal artifacts caused by iron-rich food consumption before
one of the treatment fractions compared with the artifact-free
simulation image.
5.3 Mitigation of Susceptibility Artifacts
Due to Metallic Implants
Patients with a hip prosthesis are commonly seen in radio-
5.4 Sequences for Response Monitoring
therapy departments. While many types are not a contraindi- During an MRIgRT workflow, an opportunity exists to
cation for MRI, depending on the type, some prostheses acquire sequences that may not be directly used for image
produced relatively minor artifacts, whereas others can sub- guidance. Sequences providing different image contrasts as
stantially distort the MR images. Loss of signal around the well as functional MRI may be acquired. The feasibility of
prosthesis may hamper contouring (see Fig. 5.8), particularly quantitative functional imaging has been demonstrated for
if this relies on the interpretation of diffusion images, which both the Unity and the MRIdian systems (Kooreman et al.,
44 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure 5.8. Unilateral hip replacement in a patient treated with magnetic resonance imaging (MRI)-guided radiation therapy. Axial and
coronal views (left top and bottom) of T2-weighted three-dimensional volume acquired on the Unity MRI-linear accelarator with implant
outlined in yellow, showing signal void in vicinity of implant. Axial and coronal views (right top and bottom) of B0 map acquired on the
Elekta MRI-linear accelerator, fused with planning computed tomography. The B0 field perturbations from the hip implant extend out
further than the MRI signal void and, depending on target location, may reduce spatial fidelity. Note that B0 discontinuities can be seen
due to phase wrapping artifacts in the B0 processing. Images courtesy of Joan Chick, The Royal Marsden Hospital.

Figure 5.9. Abdominal vascular stents in magnetic resonance imaging (MRI)-guided radiotherapy. Sagittal view of T2-weighted three-
dimensional volume acquired on the Unity MRI-linear accelerator (linac), showing signal void in vicinity of stents (left). Sagittal view of B0
map acquired on the Elekta MRI-linac, fused with planning computed tomography, in which stents are visible anterior to L5 (red arrow)
(right). The B0 field perturbations extend out as far as the prostate target (white arrow) and could reduce spatial fidelity. Note that
B0 discontinuities can be seen due to phase wrapping artifacts in the B0 processing. Images courtesy of Joan Chick, The Royal Marsden
Hospital.

2019; Nejad-Davarani et al., 2020; Yang et al., 2016), Incoherent Motion MRI and dynamic contrast-enhanced
although at present, most of the sequences used in publica- MRI (Thorwarth et al., 2020). As the latter technique requires
tions are not available in standard clinical operation. For administration of contrast agent, the same disadvantages
both systems, functional imagings such as T1 mapping apply that were discussed earlier.
(Mickevicius et al., 2021; Nejad-Davarani et al., 2020) and The possibility to acquire quantitative functional images
diffusion-weighted MRI (Gao et al., 2021; Lewis et al., during each treatment fraction provides a unique opportunity
2021a; Yang et al., 2016) have been demonstrated. Further for the use of quantitative imaging biomarkers (QIBs) to per-
techniques that have been used are relaxometry such as T2 sonalize treatment which may improve clinical outcomes.
and R2* mapping (Nejad-Davarani et al., 2020), techniques The report of AAPM TG-294 describes a framework to
to measure perfusion characteristics, such as IntraVoxel define imaging biomarkers and establish clinical validation,
 Imaging with an MRI-Linac 45

Figure 5.10. Simulation image without artifact (left) and daily image of the same patient with significant metal image artifacts caused by
iron-rich food consumption (right).

utility, and quantification and proposes standardized pro- Before clinical implementation, it is necessary to
cesses to achieve that (McGee et al., 2021). Recently, a road- determine the accuracy and precision of quantitative MRI
map toward the clinical use of QIBs on MRIgRT systems sequences in standardized phantom measurements
was presented (van Houdt et al., 2021a). While linking to the (Kooreman et al., 2019). Quantitative imaging biomarkers
more general imaging biomarker roadmap for cancer studies need to be accurate, repeatable, and reproducible between
(O’Connor et al., 2017), this study addresses the specific other systems. To determine if changes in a QIB observed
challenges that need to be met for effective use of QIBs during the course of treatment are significant, it is critical
within MRIgRT. The prime challenge is that the MRI sys- to measure the repeatability of the technique prior to the
tems of MRI-linacs differ considerably from diagnostic sys- start of treatment. To allow comparison of QIB values
tems. These differences can have an impact on quantitative measured during the treatment on an MRIgRT system with
sequences. Thus, simply copying sequences from diagnostic measurements done during follow-up with a diagnostic
systems may result in biased measurements. For example, system, it is important to know the reproducibility of the
for diffusion-weighted MRI on the Unity system, the b-values technique between such systems. Therefore, in the use of
(representing the strength of diffusion weighting selected for QIBs in MRIgRT, it is recommended to include the acqui-
ADC measurements) may need to be reduced compared to sition of test-retest data as well as a comparison between
diagnostic systems because of the reduced SNR and lower QIBs from MRIgRT systems and diagnostic MRI systems
gradient strength (Kooreman et al., 2020). (van Houdt et al., 2021a).
46 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

6. Online Adaptive MRIgRT

6.1 Introduction 6.2 Existing Implementations

The concept of adaptation in radiation therapy has been The superior soft-tissue contrast of MRI allows for accurate
around for over five decades (Oleĭnik and Klepper, 1973) assessment of interfractional and intrafractional anatomic
and holds the promise of personalized medicine. The general variations and creates opportunities for online adaptive
paradigm involves the identification of a clinically signifi- replanning. The currently existing commercial systems pro-
cant change, determination of the rate of that change, poten- vide functionality for online MR-guided planning adapta-
tial risk if left unmanaged, and finally, the adjustment of tion. Although the technical implementation and
radiation delivery to account for this change. There are two specifications are different across the platforms, the online
potential goals: (1) to ensure that the dose prescribed for the adaption provided by these systems is mainly based on ana-
radiation treatment is being delivered as intended even if the tomic MRI guidance and can be generally described as two
patient’s anatomy changes and (2) to optimize delivery of the basic workflows: (1) shifting the patient to optimize target-
treatment plan. This may be either by reducing the treated beam alignment analogous to X-ray IGRT, or alternatively
area (e.g., treatment margin reduction) or by continually shifting the beam to optimize beam-target alignment, with
changing the tumor dose distribution (e.g., dose escalation the latter sometimes referred to as adapt to position (ATP)
when anatomy is favorable or de-escalation due to proximity and (2) online adaption which is sometimes referred to as
of OARs to the target). However, caution must be exercised adapt to shape (ATS) (Winkel et al., 2019).
to ensure dose escalation remains in clinically acceptable Similar to conventional image guidance, daily setup MR
levels when modified online. images are rigidly registered to the planning reference images
For each rate of change, specific technologies are neces- to calculate the change in the isocenter position. In X-ray
sary for the clinical team to “keep up” and adapt to changes IGRT, the couch position (x, y, z) is usually adjusted, and
quick enough to still be useful. The first of these is being able sometimes the principal angles with respect to position (yaw,
to detect the change. The technology necessary may be as pitch, and roll) are modified as well to achieve a full 6 degree-
simple as a scale (to note patient weight changes) and as of-freedom treatment couch. Magnetic resonance imaging–
advanced as PET and functional MRI techniques that are guided radiation therapy can be performed in this manner,
also available on MRI-linacs (to note functional changes). although 6 degrees-of-freedom couches are not yet available.
The second suite of technologies necessary for adaptation For the ATP approach, instead of applying a couch shift to
includes software for image processing. This must be robust account for the change in isocenter, MLC segments from the
enough to detect differences in the status of patient’s anat- initial treatment plan can instead be adjusted based on the
omy or function relative to the initial images. Toward that new patient position, and the new plan is recalculated on the
end, image registration and anatomy segmentation are the initial planning images with the original contours. In effect,
most basic of needs. the patient is not taken to the beam but the beam taken to the
The third necessity involves the ability to quickly create patient. As no contour editing is required, the ATP workflow
an updated radiation therapy plan to achieve the prescription is relatively straightforward and efficient; however, patient-
with consideration of previously delivered fractions—again specific quality assurance of the adaptive plan may be neces-
to be able to act quickly enough to have a clinical impact. An sary due to the changes in weighting and shapes of the MLC
efficient and user-friendly workflow is necessary to ensure segments. This is often validated by independent monitor
the successful clinical implementation. The last is, of course, units (MUs) or absorbed-dose calculation programs. The
a robust method of quality assurance so as to ensure the same tradeoff of the fast ATP planning based on initial reference
safety and integrity of treatment as under standard, less- images and original contours is that it may not fully represent
changing workflows. the actual daily anatomy and often lead to violations of dose
The definitions of the different time scales of adaptive constraints when recalculating on the daily MR images
radiotherapy are shown in Table 1.1. (Winkel et al., 2019). The ATP workflow has also been
 Online Adaptive MRIgRT 47

shown to offer limited compensation for large patient offsets, segment shape and/or weighting (Tocco et al., 2020; Winkel
and different adaptation strategy variants were employed to et al., 2019). The ATS workflow is similar to the MRIdian
overcome these limitations and reduce the gap between ATP ART workflow except that dose reoptimization is directly
and full plan optimization (Gupta et al., 2022). performed following recontouring. The choice between ATP
The online workflow is essentially a full plan reoptimiza- and ATS workflow is usually made based on each institu-
tion based on the actual anatomy and adapted contours. tion’s practice protocol. The ATP and ATS workflows can
Compared with the ATP workflow, it allows adaptation for also be combined in some clinical scenarios. For example,
complicated anatomic variations such as target and OAR size the ATS workflow can be employed following the ATP pro-
changes and shape deformation at the cost of increased cess if the plan dosimetry is not acceptable with simple
resource and time requirement due to the complexity of the adjustments of segment shape and weighting. The ATP can
workflow. The contours need to be delineated based on the be used as a virtual couch correction if the pretreatment
daily anatomy, and the full replanning requires accurate elec- imaging indicates a shift of the new plan is necessary to
tron density information, fast and robust dose optimization, account for the intrafractional motion which occurs during
and calculation algorithms. the ATS process. The adaptive planning workflow for the
The online MRIgRT-adaptive workflow compresses the Aurora-RT system is conceptually similar to that of both
conventional planning process usually occurring over a few MRIdian and Unity; however, the Aurora-RT has the ability
days into a very short time span while the patient remains in to make large couch shifts in the lateral, longitudinal, and
the treatment position. It requires rigorous process manage- vertical directions; therefore, the need to “Adapt to Position”
ment and a comprehensive QA program to be discussed in is not required, and the patient can be aligned to the beam as
the next sections. planned in virtually all cases. In those cases where there are
anatomical differences and an adaptive workflow is required,
the process will be similar to what has been described for the
6.3 Workflow and Technical MRIdian system. Owing to the parallel alignment of the
Considerations for Online MRIgRT- magnetic field with the radiation beam, conventional dose
Adaptive Radiotherapy calculation algorithms can be used, and the Aurora-RT does
The generalized MRIgRT process is shown in Fig. 1.1. The not provide a dedicated planning system for its dose calcula-
clinical workflows for ART on MRIdian and Unity systems tions. Therefore, the ART workflow and the process for con-
are outlined in Fig. 6.1 and briefly described as follows. On touring and dose optimization may be slightly different
the MRIdian system, conventional image-guided patient depending on which treatment planning system is being used
setup can be performed following the daily MR image acqui- by the customer. Both online and offline adaptive workflows
sition. Couch correction is calculated and applied based on are possible. The full set of tools for adaptive planning on the
rigid registration of daily MRI with reference planning Aurora-RT system is not clinically released at this point in
images to account for the isocenter shift. Planning contours time. These workflows involve complicated processes and
are then propagated to the daily MR images through DIR fol- an interdisciplinary team, and their success heavily relies on
lowed by manual review and editing of the contours. The several key technical elements. There are additional require-
initial treatment plan can be recalculated on the daily MR ments and pressures due to the nature of planning in a highly
images with the adapted contours to ascertain whether the compressed time frame.
initial plan is still acceptable for the actual daily anatomy. If
the quantitative dosimetric assessment based on the pre- 6.3.1 Online Adaptive Treatment Planning
dicted absorbed dose leads to a clinical decision that adaptive Considerations
planning is necessary, reoptimization of the initial treatment
will be performed based on the daily MR images and con- Online MRI-based adaptive planning is essentially an MRI-
tours. The replanning can take additional iterations until an only planning process because it is impossible to acquire a
acceptable treatment plan is derived. After plan evaluation CT scan while the patient is in the treatment position imme-
and approval, patient-specific plan QA is carried out, and diately before delivery. Methods to obtain electron density
verification imaging can be acquired to validate the anatomy information include bulk density assignment, deformable
and assess the intrafractional motion before proceeding with registration with the simulation CT, and voxel-intensity-
the beam delivery. On the Unity system, ATP or ATS work- based synthetic CT as described in section 4. Bulk uniform
flow can be chosen based on the assessment of daily anatomy density assignment was found to achieve dosimetric param-
or per institution protocol. The ATP workflow is essentially a eter differences of 3 % for the PTV and 5 % for OARs com-
virtual couch shift to account for the isocenter shift calcu- pared with dose calculation using CT scans for prostate and
lated based on rigid registration of daily MR images to the pancreatic planning (Prior et al., 2016). Deformable registra-
reference images. The initial treatment plan is then modified tion of simulation CT to MRI provides a simple way to obtain
to reproduce the target coverage through adjustment of MLC patient-specific electron density information for initial
48 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Def ble
Daily MR image t Review
ec t e P edic
dE
p ga

Re- Y Need
p za Adaption

N Plan
Acceptable
Y

P e-Tx Ima
QA Tx Delive y
Valida

ATS Def ble

t Review
t
E Re- za
P ga
Y
Daily MR Image Need Full
Acq Adaption

N
Segme t N
N Plan
Rigid Regist a ape/Weig Acceptable
ATP Adjustme t
Y

P e-T
QA
Valida

Tx Delive y

Figure 6.1. General online adaptive workflows for the MRIdian (top) and Unity (bottom) systems. QA = quality assurance;
ATS = adapt to shape; ATP = adapt to position.

planning because the simulation CT and MRI are often (Pham et al., 2022). However, a much larger dosimetric per-
acquired within a very short period with negligible anatomic turbation was observed for the 1.5 T system, which necessi-
change. For online adaptive planning, this method may lead tated correct density assignment to account for the electron
to inaccurate electron density mapping due to anatomic return effect in daily plan adaptation (Shortall et al., 2021).
changes such as organ filling and air cavity variation in the Synthetic CT directly generated from the daily MR images
GI tract, as shown in Fig. 6.2. This inaccurate electron den- through voxel-based inference has the advantage of high
sity mapping requires manual modification and density over- anatomical correspondence and therefore has the potential to
rides. The dosimetric impact of air cavities was found to be further improve the accuracy of electron density mapping
minimal for prostate bed adaptive planning with the 0.35 T and dose calculation (Cusumano et al., 2020).
MRI-linac when the electron density map was directly The online ART planning consists of all the major steps
obtained from simulation CT without manual correction involved in conventional treatment planning but needs to be
 Online Adaptive MRIgRT 49

Figure 6.2. Example of inaccurate electron density mapping due to anatomical changes used for online adaptive planning. Daily
TrueFISP 0.35 T MRI acquired on a low-field MRI-linear accelerator (left). Reference electron (E) density obtained by deforming
the simulation computed tomography to the daily MRI with air in the stomach not accurately reflected (center). Density overrides
to reconfigure the air (yellow) and fill in the missing air with tissue density overrides (red) used for subsequent dose calculation
(right).

executed on a much shorter time scale. Therefore, the whole Rapid online plan adaptation requires fast and robust dose
process must be implemented in a highly efficient and robust optimization and calculation algorithms. To take the mag-
fashion. To avoid contouring from scratch, DIR is often used netic field effect into account, it also mandates Monte Carlo–
to propagate contours generated from reference images to based absorbed-dose calculation algorithms. While it used to
daily MR images. However, intensive user review and manual be very time-consuming and computationally intensive,
edits are still needed due to errors and uncertainties associated modern central processing unit and graphics processing unit
with DIR as described in section 4. Another strategy is to only hardware and parallel computation have enabled these Monte
refine a subset of the structures that are proximate to the target Carlo dose calculations to be performed in seconds (Hissoiny
volumes if the OAR tolerances are small-volume dose con- et al., 2011; Jia et al., 2011; Wang et al., 2016; Ziegenhein
straints (Bohoudi et al., 2017). In addition, derived contours et al., 2015). The inverse optimization process for replanning
such as target expansions or optimization tuning structures are is however still rather time-consuming and can easily be
often generated based on predetermined rules that can be auto- minutes. Several strategies have been developed to expedite
matically applied during online planning, which not only the replanning process while maintaining the quality and
improves efficiency but also reduces potential human opera- robustness of the plan. To minimize user selection and
tional errors. Despite these strategies to expedite recontouring, interobserver variability, beam angles and optimization
multiple clinical studies have identified contour segmentation parameters from the initial plan are often used as the starting
as one of the bottlenecks in the online ART process (Güngör point of reoptimization, but this places a higher requirement
et al., 2021; Henke et al., 2018; Lamb et al., 2017) with a for the initial planning to be robust to potential large daily
median recontour time of around 10 min. Deep-learning-based variations. Insufficient weighting on an OAR that is rela-
auto-segmentation has been investigated recently with the tively far away from the target volume in simulation may
promise to speed up the process in the order of seconds or a result in suboptimal plan quality when the OAR becomes
few minutes (Fu et al., 2018; Savenije et al., 2020). However, closer to the target during the adaptive fraction. Strategies
translation of these approaches into clinical practice still faces such as combining multiple OARs into one optimization
numerous challenges including quality of training data, inter- structure or grouping them based on distance from the PTV
institutional and observer variability, and logistical difficul- have been shown to simplify the optimization process and
ties. Another approach to improve contouring efficiency is improve the planning robustness by reducing the sensitivity
through parallel operations. Training and involvement of radi- of each organ to daily variations (Bohoudi et al., 2017;
ation therapists or other staff members to perform contouring Olberg et al., 2018). Simplified optimization methods such
has been exploited with promising results (Willigenburg et al., as segment weight optimization were found to achieve plan
2021). Recently, online planning software that allows multiple dosimetry comparable to the full-optimized plans with a sig-
users to simultaneously contour different organs has been nificant reduction of optimization time, although full-optimi-
developed which is expected to significantly accelerate and zation was still deemed necessary in some patients and
streamline the online planning workflow. fractions when large anatomic changes occurred (van
50 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Timmeren et al., 2021). Novel leaf sequencing and fluence adaptive workflow because the patient has to remain in the
optimization algorithms have been developed resulting in treatment position. At present, commercial patient-specific
shorter optimization time and faster beam delivery without QA solutions are mainly based on secondary absorbed-dose
compromising dosimetric plan quality (Snyder et al., 2022). calculation of the new adaptive treatment plan by a simpli-
Artificial intelligence–based methods have been investigated fied absorbed-dose calculation tool provided by the vendor
to help generate high-quality base plans for MRI-guided or independent in-house developed or third-party commer-
ART for locally advanced pancreatic cancers (Bohoudi et al., cial software. As the independent absorbed-dose calculation
2017). Here, 66 patient plans that met dosimetric goals were can only detect absorbed-dose calculation–related errors,
curated with dosimetric and patient-specific geometric end- conventional measurement-based QA can be retrospectively
points inputted for model training. Validation was performed performed at the end of each adaptive fraction as additional
on 42 unique patients based on patient-specific geometric validation of treatment delivery. However, it might be too
parameters, yielding robust and individualized results for late to catch significant delivery errors, and one must also
each patient. A deep learning dose calculation method has consider the practical costs associated with the increased
been shown to accelerate Monte Carlo dose calculation for workload of QA measurements of each fraction. Nevertheless,
adaptive planning by a factor of 38 over traditional calcula- it is a good practice to perform measurements for a represen-
tion (Neph et al., 2021). tative subset of patients during the initial implementation
A critical component of online ART is the clinical deci- phase to establish the correlation between measurement and
sion-making involved in assessment for adaptation and plan calculation-based methods and build confidence for the sec-
evaluation. To minimize the time and resources allocated for ondary absorbed-dose calculation (Green et al., 2019). To
the online ART process, the clinical threshold to adapt and ensure accurate delivery of the planned absorbed dose, sev-
plan evaluation criteria should be predetermined and docu- eral methods have been investigated to monitor the treatment
mented so that consistent and objective clinical decisions can delivery in real time in lieu of pretreatment QA (Glide-Hurst
be made throughout the treatment course. The adaptive et al., 2021a). Treatment machine log-file-based analysis has
directives should specify the conditions that trigger plan shown a strong correlation with measured-based QA with
adaptation, such as target coverage below a threshold and/or high sensitivity in error detection on the 1.5 T MRI-linac
violation of a stated OAR constraint. Tradeoffs between tar- system (Lim et al., 2021). When combined with the cine
get coverage and OAR sparing and priorities of different images, the machine log file can be used to potentially recon-
OARs should also be provided. It can also include instruc- struct the delivered absorbed dose taking the intrafractional
tions on recontouring essential structures and generating motion into account (Kontaxis et al., 2020; Menten et al.,
derived structures. It is recommended that consistent adap- 2020). In addition, in vivo transit dosimetry through the
tive dosimetric constraints should be developed for different Electronic Portal Image Device equipped on an MRI-linac
disease sites and treatment regimens to ensure standardiza- allows reconstruction of the absorbed-dose distribution in
tion and homogeneity in clinical practice, which is also the patient daily geometry, offering an automatic dosimetric
extremely important when developing clinical trials involv- verification method for online adaptive plans (Olaciregui-
ing ART (Glide-Hurst et al., 2021a). Owing to the time-con- Ruiz et al., 2021). Although the feasibility of these monitor-
suming and labor-intensive nature of the ATS workflow, a ing methods has been demonstrated, continuous development
clinical decision needs to be made immediately after the efforts are still warranted to deploy these techniques for reli-
daily MRI scan as to whether a full adaption is required or able real-time detection of treatment deviations to ensure
couch correction or simple beam-target alignment is suffi- safe and accurate delivery of the new adaptive plan in the
cient to account for the anatomic changes. However, this is absence of pretreatment QA measurement.
often a subjective call by the clinician who visually inspects
and compares the daily anatomy to the planning reference. A 6.4 Quality Assurance for MRIgRT-
retrospective evaluation study found that 58 % of fractions Adaptive Radiotherapy
with online nonadaptive decisions would benefit from the
full adaptive replanning, indicating quantitative metric to The increased complexity associated with the online adap-
assist in the decision-making are highly desirable. tive planning workflow may introduce new uncertainties and
Deformable image registration metrics between the planning necessitates several special quality assurance considerations
and daily images such as Jacobian determinant histogram in the initial commissioning and periodic QA processes.
have been proposed as a quantitative criterion to facilitate the Adaptive radiation therapy mandates high-quality daily
decision-making process and have shown promising results images for accurate tumor and organ delineation and
for predicting the need of treatment adaptation for prostate absorbed-dose calculation, which underscores the need for
patients (Lim et al., 2020). appropriate periodic QA of the in-room imaging system.
Conventional pretreatment measurement–based patient- Routine examination of the image performance such as geo-
specific quality assurance becomes impractical for online metric distortion, image quality, and artifacts is highly
 Online Adaptive MRIgRT 51

Figure 6.3. Accidental mouse-click during gross tumor volume editing during online adaptive planning of a bladder stereotactic body
radiation therapy case resulted in suboptimal absorbed-dose conformity as indicated by the green arrow and excessive absorbed dose
to the rectum (left). Zoomed in view shows the artifact gross tumor volume pixel with a planning target volume island generated from
automated margin expansion and overlapped with the rectum (right).

desirable for timely identification of potential imaging the AAPM TG-132 criteria and clinical acceptable end-to-
issues. Automated image QA analysis can expedite the QA end dosimetric agreement (Mittauer et al., 2020).
process to provide a fast and robust approach for frequent In addition to the end-to-end test to characterize the ART
tests to capture potential issues and ensure optimal perfor- workflow using a simulated process, failure mode and effects
mance for online ART. analysis (FMEA) has been employed to identify and quantify
Specific end-to-end tests are necessary to validate the risks for potential errors associated with online ART and to
entire adaptive planning chain and identify potential issues further develop risk mitigation strategies and allocate QA
in each step associated with different subsystems before the efforts for safe and efficient clinical implementation (Noel
clinical implementation of ART. Effective end-to-end tests et al., 2014). In an institutional analysis, a total of 89 failure
rely on an appropriate phantom that is capable of simulating modes were identified for the institution-specific online
realistic interfractional changes such as organ shape defor- adaptive workflow with 30 % of the risks related to MRI-
mation and allows multimodality imaging for both CT and specific workflows and 46 % specific to the online treatment
MRI. Several anthropomorphic deformable phantoms have planning adaption (Klüter et al., 2021). Mitigation strategies
been specifically developed for MRIgRT ART validation such as standardized workflow, clearly-defined protocols,
(Axford et al., 2021; Elter et al., 2019; Hoffmans et al., and checklists to ensure protocol adherence were imple-
2020). A comprehensive end-to-end daily QA workflow was mented to minimize all the high risks that were classified
developed to check the functionality and connectivity of all based on the combination of probability and severity levels
subsystems of an MRI-linac system for online adaptation in to an acceptable level (Klüter et al., 2021). Contouring and
a short time frame including tests on image acquisition and electron density assignment were also reported to be the
registration, data transferring, online adaptive planning as highest risk priority categories by different institutions
well as delivery and motion monitoring (Chen et al., 2020). (Garcia Schüler et al., 2021; Rippke et al., 2022). The quality
As discussed in section 4, DIR plays an important role in of the contour directly impacts the quality of the treatment
ART for contour propagation, electron density mapping, and plan as shown by an example in Fig. 6.3. In this adaptive
dose accumulation, but with a high degree of uncertainties. It planning of a bladder SBRT treatment, an accidental mouse-
is critical to develop QA methods to characterize the perfor- click during GTV contour editing resulted in a pixel of the
mance of DIR in the ART process. Deformation and dosimet- GTV contour at the border of the rectum in Fig. 6.3. After
ric accuracy of online ART on a 0.35 T system was evaluated automated margin expansion, a small artifact PTV island was
by a heterogenous anthropomorphic multimodality abdomi- generated leading to suboptimal absorbed-dose conformity
nal deformable phantom with thermoluminescent dosimeter and excessive absorbed dose to the rectum as indicated by
measurements, demonstrating deformation accuracy within the green arrow in Fig. 6.3. This kind of contouring issue is
52 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

not easily caught by the secondary absorbed-dose calculation thus, caution must be taken when applying the deformable
QA because contour and electron density are direct inputs to absorbed-dose summation in decision-making (Glide-Hurst
it, indicating the necessity to develop additional automated et al., 2021a). Another challenge associated with clinical
plan check tools to identify the issues associated with con- decision-making is that dose constraints previously estab-
touring and electron density mapping. Based on the potential lished based on nonadaptive treatment regimens may not
risks identified in the FMEA of the adaptive planning pro- reflect true OAR tolerances. Clinical studies with rigorous
cess, several in-house tools for automated contour analysis, tracking and reporting of absorbed dose to OARs are neces-
electron density examination, and additional plan checks sary to develop new adaptive absorbed-dose constraints
have been reported as complementary plan QA to the sec- accounting for daily anatomic changes.
ondary absorbed-dose calculation (Cai et al., 2018; Rippke Intrafractional motion during the online ART process
et al., 2022). remains a major concern because the adaptive plan may no
longer be relevant to the anatomy if substantial anatomic
changes such as bulk or peristaltic motions occur during the
6.5 Challenges for MRIgRT ART
planning process. For example, it was reported that a 6-mm
In addition to the workflow and QA challenges discussed planning margin was required to compensate for the
above for the safe and robust implementation of MRIgRT intrafraction GTV motion occurring during the online ART
ART, several technical challenges must be tackled to maxi- process for rectal tumors treated on a 1.5 T MRI-linac
mize the potential clinical benefits of this technology. (Eijkelenkamp et al., 2021). The margin can be reduced to
As a new treatment plan may be generated for every frac- 4 mm, if the procedure time can be controlled within 15
tion with varying anatomy in ART, it is very challenging to min. Active intervention such as predelivery volumetric or
estimate the actual cumulative absorbed dose, accounting cine imaging can be used to detect the intrafractional
for both anatomic and dosimetric changes. In current clini- motion followed by couch correction or additional adaptive
cal practice, the conservative isotoxicity approach is com- planning. Emerging real-time ART techniques such as
monly employed in which each new adaptive fraction is treatment gating, dynamic MLC tracking, and real-time
evaluated de novo as if it would be delivered for the entire absorbed-dose calculation and replanning (Kontaxis et al.,
treatment course, and OAR sparing is prioritized over target 2017) have been introduced and extensive technical devel-
coverage (Green et al., 2019). To simplify the plan evalua- opment is still needed to fully implement these techniques
tion and clinical decision-making, the near-maximum into clinical practice.
absorbed dose or absorbed dose to a small volume of the Biologically guided ART holds great promise because
OAR is summed over treatment fractions, rather than con- changes at physiologic and molecular levels characterize the
sidering the actual delivered volumetric absorbed dose to underlying biological response to radiation treatment and
the OAR and the day-by-day location of the maximum may have prognostic value with potential to modify treat-
absorbed dose. This conservative approach of direct ment (van Houdt et al., 2021b). Recent studies have shown
absorbed-dose summation without considering its spatial that QIB measurement is feasible on MRI-linac systems
location is simple to implement within the compressed time even with a low magnetic field (Nejad-Davarani et al., 2020;
frame of online ART, but it can also lead to an unnecessary Yang et al., 2016). A major challenge is to develop a rigorous
sacrifice in target coverage and is only applicable when the QA program to ensure robust measurement with high repeat-
maximum OAR absorbed-dose constraints are clinically rel- ability and reproducibility. Another challenge is to develop
evant, such as for serial-like normal tissue architectures, and strategies and methods to incorporate biological information
can be determined a priori. An alternative approach to obtain into the decision-making of the adaptive process including
cumulative delivered absorbed dose for total fractions to bioeffect modeling, for example, equieffective dose
date is through voxel-by-voxel dose summation by deform- (EQDXα/β) and when and how to adapt. Tools for image
ing the dose of each fraction based on DIR of the daily interpretation and assessment need to be developed to assist
images to initial planning images (Lim et al., 2014; Veiga in the fast online decision-making process and strategies for
et al., 2014). However, the accuracy of each cumulative integration with anatomic adaptation also need to be estab-
voxel absorbed dose highly depends on the DIR algorithm, lished. Clinical studies investigating the necessity of online
the magnitude of anatomic changes, and underlying image biologically guided ART are also warranted to demonstrate
quality which often is associated with large uncertainties, the clinical value of this approach.
 MRIgRT Staffing, Training, and Safety 53

7. MRIgRT Staffing, Training, and Safety

Each department must define local rules and responsibilities members. For a more traditional IGRT workflow with an
according to national recommendations. American College MRI-linac, some centers routinely operate with just two radia-
of Radiology recommendations (ACR, 2020) often serve as tion therapists (Hales et al., 2020) and bring in additional team
a guideline to the safety hierarchy including several key per- members for the more complex adaptive or gated patients.
sonnel which can be extended to a radiation oncology setting Expert radiology input may be needed, at least during
for MRI-linacs in the following manner: (1) the MR Medical early clinical experience where MRI is not part of the normal
Director (MRMD), or a licensed radiation oncologist/radiol- planning pathway. Roles will differ internationally: in some
ogist with appropriate training in MR safety who assumes countries, for example, Australia, Canada, and the
overall and ultimate responsibility for MR facility opera- Netherlands, radiation therapists are experienced treatment
tional safety, (2) the MR Safety Officer (MRSO), or a trained planners and may contour targets and OARs and create plans.
individual, often an MR technologist, medical physicist, or Discussion to train and elevate the role of a radiation thera-
radiation therapist, who is responsible and should oversee pist to an advanced radiation therapist is underway in many
safety practices within a defined component of the MRI countries. To reduce the resources and staff involved, roles
practice at all times, and (3) the MR Safety Expert (MRSE) and responsibilities will need to change and evolve. Need
who serves as a resource to the MRMD and MRSO for non- may arise for increased reliance on automation, for example,
medically related MR safety issues (i.e., issues other than automatic segmentation and using the absorbed-dose con-
contrast agents, anxiolytics, and other pharmaceuticals). The straints as the benchmark for definitive plan approval.
MRSE is often an MRI physicist or has equivalent expertise However, some human interaction is likely to remain, and it
and may be external to the organization, whereas the MRMD will be important for responsibilities and thresholds of
and MRSO are assumed to be part of the organization per- acceptability to be well defined within the multidisciplinary
forming the MRI procedure. Organizations should collabo- team.
rate closely with their hospital and radiology colleagues to
establish proper policies and procedures related to the MRI
7.2 Training
safety program in radiation oncology. Each staff member
should be aware of potential hazards and work safely within In implementing an MRIgRT program, hospital management
their authorization level and designation. must be aware that all staff working on these systems must
be adequately trained. It must be emphasized that the typical
7.1 Staff Requirements to Start and training programs for radiation therapy professionals will
often not satisfy the requirements for working on an MRIgRT
Maintain a Program
system. In many jurisdictions, operation of an MRI scanner
A multidisciplinary team with dedicated expertise and respon- is a restricted practice, and only those with appropriate certi-
sibility for MRIgRT is required, including radiation therapists fication and licensure are able to operate them. This will
(McNair et al., 2021), clinicians, physicists, and treatment often restrict the staffing requirements on the MRIgRT sys-
planners. Staffing requirements will initially be resource tem to require radiation therapists to be dual certified, or the
intensive with professionals performing traditional roles but machine will need to be staffed by a radiation therapist and
within a much shorter time span and under time pressure. For an MRI technologist working together collaboratively. Often,
example, radiation therapists positioning patients, acquiring MRIgRT programs will have a qualified radiation therapy
and registering images, and delivering the treatment is part of physicist working alongside a qualified magnetic resonance
routine practice. However, having clinicians contouring tar- imaging physicist. While this can work well, both individu-
gets and OARs live and physicists or treatment planners creat- als will need adequate cross-training to ensure they are both
ing and checking plans in real time requires dedicated and aware of all safety features of the device. Although the prac-
protected time. Alternatively, many clinics have implemented tice of the radiation oncologist has potentially included using
a treatment team of the day dedicated to this process. As in MRI in their planning workflow, it must be recognized that
conventional radiotherapy, the complexity of the treatment not all radiation oncologists will be comfortable with using
will dictate the level of involvement of the above team MRI in the treatment environment without further training.
54 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Table 7.1. Summary of additional training requirements by profession.

MRIgRT role

Radiation
Radiation MRI Treatment therapy MRI Radiation
Additional training required therapist technologist planners physicist physicist oncologist
MR safety   
Radiation safety  
Radiation therapy physics 
MR imaging   
MR physics (pulse sequence and image weighting)    
Causes of MR imaging artifacts and noise sources  
MR quality assurance  
MR anatomy and target contouring     
MR image matching/fusion    
RT immobilization devices 
RT quality assurance 
Target volume definitions  
Radiation dosimetry  
MRI-only-based planning    
Adaptive planning      
Image-based gating      

Regardless of the profession, all individuals that will be use of MR imaging in their practice may already be com-
entering the MRI environment will need to be trained on the monplace, and they would already have some of the skills
safety issues and screening to protect themselves, the patient, set out in the table; however, those that seldom use MRI for
and the equipment, as significant injury or even death can be planning would need to update their skills according to the
caused by bringing unsafe items such as scalpels, scissors, table below. This table should be considered a reference;
and screwdrivers into an MRI room. This also includes however, encouraging all staff to become more familiar with
implants or devices that may not be MR safe. all aspects would be beneficial.
Beyond the safety aspects, the practice of radiation ther- It is important to have adequate resources available to the
apy will significantly change when implementing an MRIgRT team to ensure a safe environment. An individual
MRIgRT program. In many cases, the introduction of well trained and experienced in MRI safety should be
MRIgRT program will bring with it the introduction of more assigned as the MRSO. This would typically be a qualified
advanced treatment protocols such as adaptive therapy and MRI medical physicist, a senior MRI technologist, or an
image-based gating. Special training in these techniques MRI radiologist. The MRSO should be responsible for train-
must be included in the development of the MRIgRT pro- ing all staff in the specific MRI safety considerations for
gram. In many cases, the scope of practice is changing as each MRI-linac installation and authorize staff to access the
these advanced techniques are being used. In some situa- area. The MRSO should be consulted if there is ever any
tions, the radiation therapists are being given broader roles question about MRI safety. However, this individual should
with online target contouring and adaptive planning than also be aware of the general radiation safety considerations
they had when all treatment planning was done by treatment as well to ensure that both are adequately considered in all
planners and physicists. Although dependent on jurisdic- situations. Similarly, the radiation safety officer (RSO) needs
tions and regional regulations, guidance on the educational to be aware of the MR safety considerations as well because
requirements of professionals and teams working in this will often be the first MR installation in the radiation
MRIgRT are emerging, such as the UK’s IPEM Topical therapy department.
Report, Guidance on the use of MRI for external beam There are five main sources of hazards in the MRI
radiotherapy treatment planning (Speight et al., 2021a) and environment:
the UK’s Society and College of Radiographers guidance,
MRI Guided Radiotherapy (SCoR, 2020). Table 7.1 sum- •• Static magnetic field: The static magnetic field can
marizes the additional training and experience that each of pull any ferromagnetic object toward the magnet with
the radiation professionals should achieve before operating significant force thus creating a projectile. The main
an MRIgRT system independently. For some clinicians, the magnetic field can also interfere with electrical
 MRIgRT Staffing, Training, and Safety 55

devices or wipe the magnetic strip on credit cards. Of entering the MRI room. The Elekta Unity has a screening
particular concern are implanted medical devices such checklist integrated into its workflow. The ACR has defined
as pacemakers, implanted cardiac defibrillators, glu- four safety zones (ACR, 2020) that should be considered in
cose monitors, and so on, that may have their opera- all MR facilities. These same zones should be applied to
tion affected by the magnetic field. The magnetic field MRIgRT facilities as well. Any visitors/staff that need to
may also cause torque on an implant such as an aneu- enter Zone IV will need to complete and sign a checklist also.
rysm clip that could cause serious complications. No It is important to do a verbal confirmation with the patient/
patient or visitor should be allowed inside the 0.5 mT volunteer. Having had an MRI scan previously does not
(5 Gauss) line without having been screened for such guarantee subsequent safe examinations because of varying
devices. field strengths of the magnet. Each day, the patient must be
•• Time varying magnetic fields: The MR gradients will confirmed to be safe to scan prior to bringing them into the
generate time varying magnetic fields, which can treatment area. The radiation therapist/MR technologist must
cause peripheral nerve stimulation. Attention should take the role of the gate keeper to ensure that proper proce-
be paid to setting up the patient to avoid loops caused dures are followed prior to entering the room. These indi-
by clasping hands, and so on. These time varying viduals should be aware of the boundaries between Zones I,
magnetic fields can also cause interference with II, III, and IV, and they must follow the local SOP to ensure
patient monitoring equipment. only appropriately screened people are allowed into each
•• Radiofrequency fields: The RF power from the trans- zone. The boundary of the 0.5 mT line should be marked on
mit coil can cause local heating and possible RF burns. the floor of the room, to ensure that all staff know where this
Attention should be paid to the specific absorption boundary is. In the case of MRIgRT systems with a rotating
rate (SAR) to avoid detrimental effects of local heat- MRI magnet (e.g., MagnetTx Aurora-RT system), the 0.5
ing. Stray RF from noise sources in the room may mT line can vary with gantry angle, and as such should be
cause image artifacts. This is especially relevant in the marked as the worst-case scenario.
MRI-linac environment where the additional equip- Devices and objects can be categorized as MRI safe, MR
ment required for radiation therapy delivery can be a conditional, and MRI unsafe. No MRI unsafe objects can be
source of such noise. brought into the MRI environment. Items that are marked
•• Acoustic noise: MRI scanners will produce a signifi- MR conditional which is defined as the device being able to
cant amount of noise during image acquisition. It is be safely scanned with specific MR hardware (i.e., maxi-
important that the patient has adequate hearing pro- mum B0 field strength, maximum spatial gradient of the B0
tection during their treatment. This needs to be con- field, and maximum SAR, etc). It is up to the local staff to
sidered at the time of making immobilization shells to determine if the MR conditional item is safe to be brought
ensure the hearing protection can be used. into the room based on an evaluation of the local conditions
•• Liquid helium filling/quenches: MRI scanners from compared to the conditions in which it was tested. It is rec-
Elekta and ViewRay are cooled by liquid helium. This ommended that the MRSO keeps a list of all MRI safe, and
leads to hazards to the service team that must replen- MRI conditional devices which have been approved for use
ish the liquid helium periodically. It also means that in the MRI environment.
there is a possibility of the magnet quenching. Quench Some implantable medical devices are marked as MR
pipes are provided in these installations to mitigate the Conditional. It is important to analyze both the MRI condi-
hazard related to quenches, but all staff need to be tions as well as the radiation safety aspects for implanted
aware of the possibility and understand the proper devices such as pacemakers or defibrillators. Devices such as
procedures to follow in the case of a quench. continuous glucose monitoring systems are now common-
place and should be assessed for both radiation and magnetic
safety before entering the Zone IV area. Available public
7.3 MRI Screening databases such as http://www.mrisafety.com and http://www.
A local standard operating procedure (SOP) should be devel- magresource.com contain MRI safety information for many
oped by each MRIgRT facility to determine a procedure for implanted devices. While implanted devices may have the
screening everyone prior to entering the MRIgRT room. designation of MR conditional, this does not equate to
Although this can largely be the same as for a diagnostic patients who can be imaged at all field strengths as many
MRI facility, it should account for the fact that some patients testing conditions are at 1.5 and 3.0 T which may not apply
may be getting daily treatments for as much as 8 weeks. A to the 0.35 T MRIdian and 0.5 T Aurora-RT. The MR staff
process for verifying any changes since the original screen- will be required to determine under what conditions patients
ing should be included. A screening checklist (examples are with MR conditional devices can be safely imaged. Factors
shown in Appendix B) should be developed and must be including the maximum spatial gradient of the static mag-
completed and signed by all patients/volunteers prior to netic field, whole body gradient systems maximum slew rate
56 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

per axis, and maximum SAR for the acquisition conditions these cases, it is important to use ear plugs, but the
(e.g., whole body or head) should be compared to that for the patient should have a call button to allow them to
implanted device. The final decision for scanning shall be alert the operator if they are in distress.
based on the individual risk-benefit decision made by an 3. Proper procedures will need to be put in place to
attending physician (qualified radiologist or radiation oncol- ensure that the radiation vault has been cleared prior
ogist) who has level 2 safety training and the patient. Another to beaming on. As some MRI-linac designs use false
consideration for treating patients with implantable cardio- walls or equipment rooms within the vault, these
verter defibrillators on MRI-linacs is the availability of addi- areas must be properly controlled. This is typically
tional resources such as for device reprogramming or achieved by placing “Last Person Out” systems in
interrogation and monitoring by electrophysiology during these areas along with video monitoring.
daily treatment. 4. Items such as closed-circuit televisions and inter-
coms must be MRI compatible.
7.4 Radiation Safety As with any linac, it will need to meet the local require-
Any facility that operates a linac should have a formal radia- ments to be licensed or registered with the radiation protec-
tion safety program with assigned responsibilities to an RSO. tion governing bodies. The amount of radiation shielding
From a radiation safety perspective, most MRIgRT facilities will be quite different between the various devices, as some
will be similar to conventional linac facilities with a few key have integrated beam stops, and the scattered radiation pro-
differences. files will be very different for each design. Once installed, all
facilities should have a full radiation survey to ensure the
1. With the design of MRI-linacs, it can be difficult to safety of staff and public outside of the treatment room. This
visually monitor the patient on a closed-circuit should be done by a qualified medical physicist in conjunc-
television. tion with the RSO. Particular attention should be paid to any
2. Audible communication with the patient will be more parts of the barrier between the radiation area and the acces-
difficult compared to a conventional linac, since sible area that needed to be penetrated for items such as the
hearing protection will be mandatory. Typically, an quench pipe or seams in the wall or ceiling where openings
MR intercom system uses headphones but that may in the barrier were created to allow the machine to be brought
not be possible with some immobilization devices. In into the vault.
 MRIgRT Commissioning and Dosimetry 57

8. MRIgRT Commissioning and Dosimetry

Given the unique components of the MRIgRT systems as has led to the development of magnetic field compatible ver-
described in section 3 and summarized in Table 3.1, includ- sions for 2D and 3D electronic detectors (de Vries et al.,
ing the presence of a strong magnetic field, delivering radia- 2018; Desai et al., 2021; Houweling et al., 2016; Mönnich
tion through components of the MRI system but also the et al., 2020; Yadav et al., 2020). With these tools, similar QA
absence of a light field, restricted access around the isocen- measurements as for conventional linacs can be performed
ter, nonrotating collimator, and solely co-planar delivery, for commissioning and acceptance procedures (Roberts
modifications to conventional QA workflows are required. et al., 2021). The magnetic field can also affect the dose
For example, designing alternate workflows and using dedi- response of detectors as described in section 8.4.
cated or altered equipment may be necessary as the presence Restricted access around isocenter and table motion: The
of the magnetic field may alter the dose distributions and MRI system limits access to the joint linac and MRI isocen-
affect QA equipment. ter position, and it also limits the degrees of freedom for
table motion. For the Elekta Unity system only cranial-cau-
8.1 MRI-Linac Commissioning and dal table motion is possible, for the ViewRay MRIdian sys-
tem, motion in all three directions is possible, with a
Ongoing QA maximum lateral motion of 7 cm. This impacts the setup of
At present, formal guidance and tolerances set by multi-insti- QA equipment similarly as the absence of the light field.
tutional end users that are highly specific to MRI-linac- Moreover, standard water tanks typically do not fit in the
specific QA programs are still under development. bore of the MRI system, and standard driving motors typi-
Nevertheless, given the existing well-established standards cally do not operate in the presence of a magnetic field (Smit
for linacs, diagnostic MRI, and MR simulators, many of the et al., 2014). Therefore, alternative solutions have been
principles can be applied to commissioning an MRI-linac as developed. Dedicated watertanks are now available. The
outlined in Table 8.1 with key endpoints and relevant AAPM setup of dosimetry equipment can be done off-isocenter and
Task Group guidance. Typical safety and image quality tests then table motion is applied to move the setup to isocenter.
of MRI relevant to the use for MRIgRT are outlined in Tables For the Elekta system a setup jig is used to facilitate position-
VI, VII, and VIII in AAPM Task Group 284 (Glide-Hurst ing and the exact position can be measured with the inte-
et al., 2021b). Consensus papers describing QA processes grated Megavoltage (MV) imager.
and workflows from multiple MRIgRT end users have also Nonrotating collimator: In the current clinical configura-
been published for the Elekta MRI-linac consortium for tions, the collimator cannot rotate, and as such, it cannot be
machine QA with tolerances specified (Roberts et al., 2021). used to define the mechanical or radiation isocenter as is
For the linac-specific QA, many tests and QA procedures typically done in conventional QA procedures (Kirby et al.,
are similar to conventional linacs and are extensively described 2006; Klein et al., 2009). A new method is required to
and used in the medical physics radiotherapy community, for achieve this with a nonrotating collimator. One such test that
instance in AAPM TG-142 (Klein et al., 2009) and Kirby et al. has been devised includes using projection images of a ball
(2006). These present the starting point for the procedures and bearing phantom from different gantry positions on the
relevant parameters to be checked. While the goal of the QA Elekta Unity system using the integrated MV imager
tests is the same, adaptations are required to mitigate the (Woodings et al., 2021). For the ViewRay MRIdian system,
impact of the differences with conventional linacs, the impact film-based approaches as well as multiaxis ionization cham-
of the most striking differences are as follows. bers arrays are in use (Latifi et al., 2019).
Magnetic field: The presence of the magnetic field For the MRI scanner, an additional set of QA actions is
impacts the safety of handling QA equipment with ferromag- required for the MRI, which are again similar to the existing
netic components, including trolleys and tools. Personnel QA procedures for diagnostic MRI scanners. The QA proce-
need dedicated training for working safely in a strong mag- dures are tailored to the radiotherapy specifications, for
netic field environment, like that of a diagnostic MRI envi- instance on spatial integrity and geometrical stability as
ronment, as described in section 7. In addition, the described by Tijssen et al. (2019) and AAPM TG-284 (Glide-
performance of many electronic detectors is affected. This Hurst et al., 2021b).
58 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Table 8.1. Example monthly QA tests for MRIgRT systems and relevant references.

Category Endpoint Key reference(s)

Mechanical Couch position/repositioning indicator accuracy AAPM TG-142 (Klein et al., 2009) and AAPM
Couch repositioning TG-66 (Mutic et al., 2003)
Localizing lasers
Gantry angle indicator
Absolute MLC position accuracy
Safety Audio/visual monitoring functionality AAPM TG-284 (Glide-Hurst et al., 2021b)
In patient alert (squeeze ball)
Compressor
Helium pressure
Dielectric gas reading
Beam-on indicator AAPM TG-142 (Klein et al., 2009)
Beam pause
Door lock functionality
Couch retraction
Radiation Flatness/symmetry AAPM TG-142 (Klein et al., 2009)
Output: stationary and gating
Field size
Imaging Uniformity AAPM TG-284 (Glide-Hurst et al., 2021b),
Spatial integrity AAPM Report 100 (Huq et al., 2016), ACR
Image quality (high/low-contrast resolution, percent signal ghosting) MRI Quality Control Manual (ACR, 2015),
Coil functionality and ACR Large Phantom Test Guidance
(ACR, 2018)
End-to-end MR/laser isocenter coincidence AAPM TG-142 (Klein et al., 2009)
testing MR/radiation isocenter coincidence
Note. Overall, MRI-linac QA can be divided into three major sections: linac-specific QA, MRI-specific QA, and QA on the MRI-linac specifications.
QA = quality assurance; MLC = multileaf collimator; ACR = American College of Radiology.

8.2 Reference Dosimetry: Beam verification may be performed by a notified body or refer-
Quality Specification ence laboratory such as the Imaging and Radiation Oncology
Core (IROC) at MD Anderson Cancer Center or the
Efforts are currently underway to standardize reference Radiotherapy Trials Quality Assurance (RTTQA) in the
dosimetry for MRI-linacs, with each major platform having United Kingdom.
unique considerations. The beam quality specifier typically
implemented in AAPM TG-51 (PDD10,x) is not practical to 8.3 Dosimetry Equipment and
measure given the large SAD for the Unity system and the
Requirements
presence of the ERE. Thus, for reference dosimetry, many
Unity end-users are implementing IAEA (2001) TRS-398 When taking dosimetry measurements in the presence of
coupled with a modified formalism (O’Brien et al., 2016) magnetic fields, several considerations should be made for
presented in the literature using TPR20,10 as the beam quality the MR-safe/conditional conditions, orientation of device
specifier which is also less susceptible to magnetic field per- with respect to the magnetic field, introduction of RF inter-
turbations (Malkov and Rogers, 2018). Many Elekta MRI- ference, and the impact of the magnetic field on the dosime-
linacs are calibrated to yield an output of 1 cGy/MU at SAD ter performance. Some devices, such as thermoluminescent
and a depth of 5 cm for a 10 cm × 10 cm field size. For dosimeters (Steinmann et al., 2019) and diode arrays (de
ViewRay end users, on the contrary, the beam quality speci- Vries et al., 2018; Houweling et al., 2016) have shown mini-
fier is the TMR(10, 10 × 10 cm2) for an SAD setup with mal changes in performance in the presence of magnetic
output at the depth of dose maximum (e.g., 1.5 cm) for an ion fields. Radiochromic film has been shown to have varied
chamber at 90 cm SAD equal to 1 cGy/MU. All machine effects in the magnetic field based on the make and model.
output calibration shall be performed with an appropriate For example, previous generations of film such as Gafchromic
MRI-compatible ionization chamber in a water tank with EBT 2 (Ashland, Inc, Covington, KY) evaluation in the
routine (e.g., monthly) constancy checks performed in water MRI-60Co MRIgRT system showed an impact on the crystal
or water-equivalent plastic with air gaps properly managed. orientation and polymerization in the active layer of the film,
As with conventional linacs, independent absorbed-dose leading to a net optical density increase of 4.3 % to 8.7 %
 MRIgRT Commissioning and Dosimetry 59

depending on the channel of interest with under-response and 1.5 T), highlighting that chamber response was more
also reported (Reynoso et al., 2016). Gafchromic EBT 3 impacted by asymmetric air gaps (~1.6 %) than by symmetric
film, on the contrary, has shown negligible differences in air gaps (<0.5 %) with the chamber parallel to the magnetic
agreement of beam profiles with and without 0.35 T mag- field (O’Brien and Sawakuchi, 2017). The angular depen-
netic fields for photons (Steinmann et al., 2019), thus spe- dence of chamber response in the presence of air between the
cific use cases and types of film should be characterized chamber walls and phantom inserts has also been evaluated,
based on the desired level of accuracy in the magnetic field. where the magnetic field interaction with the secondary elec-
trons in air reduced the collected charge by ~1 % for the
Unity MRI-linac (Hackett et al., 2016).
8.4 Impact on Ionization Chamber
Performance
8.5 Phantoms
Owing to the presence of air in ionization chambers, the
magnetic field has had a more marked effect where detector Several unique phantoms are integral to the quality assur-
performance and the need for a correction factor depends on ance of MRIgRT programs as also outlined in section 7. For
chamber orientation, air cavity geometry/volume, and known routine magnetic field homogeneity, a large (>35 cm diam-
internal geometries, which have all been included in several eter) homogeneous spherical phantom is often used. To eval-
review articles (de Pooter et al., 2021; O’Brien et al., 2016). uate large FOV spatial distortion and residual GNL distortions
The ionization tracks in matter, rather prominent in the air after 3D vendor corrections have been enabled, a large FOV
cavity of air-filled ionization chambers, will change due to phantom (>80 % of the useable FOV) with embedded land-
the Lorentz force which affects the charge collection marks or grid is recommended (Glide-Hurst et al., 2021b).
(Meijsing et al., 2009). This depends on the shape of the ion- Many MRIgRT programs rely on vendor-provided phantoms
ization chamber, as well as the orientation relative to the for ongoing QA and incorporate standard MRI testing for
magnetic field and incident beam. This effect can be miti- image quality constancy (e.g., image uniformity, contrast,
gated by chamber-specific magnetic field correction factors resolution, and ghosting) following established QA stan-
(Glide-Hurst et al., 2021b). A comprehensive set of Monte dards for MRI such as using the ACR Large Phantom.
Carlo simulations for magnetic field dosimetry was con-
ducted including 32 cylindrical and three parallel-plate ion 8.6 Motion/Adaptation
chambers to evaluate the beam quality and angular depen-
dence of these factors and quantify the magnetic field effects The AAPM TG-76 report (Keall et al., 2006) provides exten-
on %dd(10)x and TPR20,10 (Malkov and Rogers, 2018; 2019). sive guidance on respiratory motion management. In terms
Another point of attention with detector performance in of MRIgRT, key considerations including benchmarking
the presence of magnetic fields is the potential dose-response motion estimates using a programmable MRI-compatible
dependence on chamber orientation relative to the magnetic dynamic motion phantom that can be used to simulate human
field. Again, this is most notable with air-filled ionization breathing conditions with excursions >5 mm (e.g., the
chambers and must be considered for instance in reference threshold for motion management). For MRI-linacs that are
dosimetry (de Pooter et al., 2021). Also, thin air layers equipped with beam gating functionality based on imaging
around detectors can perturb the readings and should be pre- feedback, AAPM TG-142 guidance should be followed for
vented. Conventionally, such point detectors for scanning interlock testing (functional), beam energy constancy (dose
dose profiles are often used in combination with a water tank delivery to a moving target within 2 % of stationary), and
(Smit et al., 2014). In a comprehensive evaluation of the temporal accuracy of phase/amplitude gate on (e.g., latency
Unity MRI-linac, angular dependence was noted on four ref- to be within 100 ms of expected value). Measuring beam-on
erence class chambers, with the most notable effect (~5 % latency may require specialty equipment and the help of ser-
increase) occurring when larger volume chamber axes were vice personnel to measure beam-on events with an oscillo-
perpendicular to the direction of the magnetic field (e.g., tip scope. In the MRIdian platform, where gated deliveries are
pointing in the same direction as the Lorentz force) possible, the motion of the intended target is evaluated for
(Iakovenko et al., 2020). A comprehensive set of simulations each frame of a cine-MRI against predefined parameters (tar-
has been performed to characterize the complex dependence get contained within a predefined boundary) and percentage
of detector response on factors such as magnetic field of target that is out of the prescribed boundary.
strength, field orientation, chamber radius, and internal
chamber variations (i.e., impact of chamber dead volume 8.7 Surface and Exit Dosimetry
adjacent to the guard electrode where dose is not scored)
(Spindeldreier et al., 2017). In addition to chamber depen- Given the secondary ERE that occurs between dense materi-
dencies, Monte Carlo simulations have further illuminated als and less dense materials (e.g., air) coupled with the lack
the impact of air gaps on Farmer-type chamber placed in a of electronic equilibrium, surface dosimetry in MRI-linacs
water phantom across several field strengths (0, 0.35, 1.0, has been extensively modeled using Monte Carlo techniques.
60 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure 8.1. Monte Carlo simulations demonstrating the impact of magnetic fields on electron transport in air, near entry and exit surfaces
and in-phantom for a perpendicular magnetic resonance imaging linear accelerator (MRI-linac; left) and an inline MRI-linac (right). The photon
beam edges are shown in green. The electron paths are in red. Positron paths are shown in blue. From Oborn and Burigo (2021).

Significant skin doses have been revealed by simulations in-phantom for MRI-linacs are displayed in Fig. 8.1 with the
with stronger magnetic fields and large positive surface electron trajectories shown in red.
angles due to the Lorentz-force perturbation of secondary
electrons (Oborn et al., 2010). Simulations of the ESE have
8.8 End-to-End Testing
also been conducted using an inclined phantom where maxi-
mum doses increased ~38 % to 39 % and 30 % to 36 % at 0.35 End-to-end testing or the evaluation of uncertainties in the
and 1.5 T, respectively, for angles of 30° to 45°. However, localization and treatment delivery process is a necessary
this dose decreased quickly to <2 %, a rough convergence step before clinical release of the system. In an MRI-linac
between the different inclines, and a drop below 2 % in the setting, considerations include using a phantom setup con-
magnetic field simulations is attained at a depth of about taining materials that produce signal in MRI and known reg-
1 cm into the water phantom (Malkov et al., 2019). This phe- istration landmarks, incorporating a dosimeter where the
nomenon has been most pronounced in protruding anatomy magnetic field has minimal or well-characterized impact on
perpendicular to the magnetic field due to spiraling contami- the measurement, and ideally using a phantom with hetero-
nant electrons, with film dosimetry experiments revealing an geneities present to evaluate the impact of MRI-linac-specific
increase of ~6 % of the maximum absorbed dose on the cen- components such as the ERE, electron density assignments,
tral axis for a 10 × 10 cm2 beam (7 MV, 1.5 T) 5 cm outside and possibly for benchmarking online adaptive radiation
of the field edge (Hackett et al., 2018). For exit dosimetry, therapy. Some institutions have developed in-house end-to-
introducing an exit bolus has shown to reduce the patient end phantoms to evaluate the entire MRI-linac workflow
absorbed dose from the ERE. Surface and exit dosimetry have including MRI simulation, dose calculation, IGRT localiza-
also been evaluated using many measurement techniques tion, and beam delivery (Singhrao et al., 2020). One such
including film, parallel-plate chambers (Hackett et al., 2018), example is the anthropomorphic pelvic end-to-end phantom
and optically stimulated luminescence dosimeters (Kim et al., (Cunningham et al., 2019) with rectal/bladder status chang-
2020a). Measurements conducted on a 1.5 T MRI-linac with ing capabilities that was used to benchmark an MRI-linac
optically stimulated luminescence dosimeters showed that the ART workflow. Anthropomorphic phantoms are also avail-
TPS effectively modeled the ERE near the exit surfaces of a able for MRI-linac end-to-end testing from IROC with
phantom setup. Confirmation between the measured and mod- implanted film and thermoluminescent dosimeters to per-
eled surface and exit dosimetry values should be performed at form independent verification of the entire treatment plan-
time of TPS commissioning. The impact of magnetic fields on ning and delivery process (Steinmann et al., 2020) or via the
electron transport in air, near entry and exit surfaces and RTTQA credentialing process in the United Kingdom.
 Summary, Recommendations, and Future Outlook 61

9 Summary, Recommendations, and Future

Outlook

9.1 Summary geographic regions. Individual centers should therefore

agree with key stakeholders on how best to adopt the rec-
MRI-linacs are a rapidly emerging technology for radiation ommendations in the context of their MRIgRT goals and
therapy that has undergone strong clinical adoption. needs.
Magnetic resonance imaging-linear accelerators have
higher capital, operating, staff, and workflow costs than
conventional X-ray-guided linacs. However, MRI-linacs
9.2.1 New MRIgRT Adopters
provide the radiation treatment team improved soft-tissue
discrimination and the ability to monitor patient anatomy 1. Work with the hospital management team to ensure
prior to and during treatment without additional ionizing the appropriate resources are in place to ensure the
radiation risks. This additional information introduces MRIgRT program is safe and viable.
many opportunities to improve patient outcomes. Having 2. Organize into a dedicated multidisciplinary radiation
improved tumor visibility obviates the need for implanted therapy, radiation oncology, and medical physics
fiducials with subsequent improvements in tumor targeting. MRIgRT team, ensuring sufficient redundancy for
The safety (CTV-PTV) margins can be reduced, thereby staff turnover and leave.
reducing the likelihood and potential severity of treatment- 3. Invest in education and training: Discipline-specific
induced side effects. Magnetic resonance imaging-linear specialist training is required for radiation therapists,
accelerators, particularly when coupled with online daily radiation oncologists, and medical physicists, includ-
ART, have also enabled opportunities for hypofractionation ing MRI safety training and MRI-based contouring
(e.g., completing the treatment course in fewer sessions), training for clinicians where this was previously
thereby mitigating the cost and time commitment for the CT-based.
patient and the health care system. The traditional three- 4. Use available resources as a starting point for knowl-
step image-plan-treat workflow in radiotherapy can be inte- edge accumulation and retention (as summarized in
grated into a single treatment session further improving Appendix A).
patient convenience and costs. As MRI-linacs combine two 5. Initial treatments at new centers may start with lower
devices with their own risk profiles, and they are an emerg- risk MRIgRT procedures such as palliative cases,
ing technology, the safe implementation requires skilled pelvic sites, and standard fractionation cases where
teams, ongoing training, a safety-first culture, comprehen- motion is limited.
sive QA programs, and clearly documented treatment 6. Perform end-to-end tests for new treatment sites.
workflow procedures. As experience grows and clinical 7. Seek to standardize clinical processes, while under-
evidence builds for MRIgRT, identifying the most suitable standing that there is significantly more complexity
populations and quantifying the benefits of this technology and more potential options in various aspects of the
will be fully realized. system and particularly imaging sequences that can
be personalized.
8. The aim of online adaptation is to reproduce the
9.2 Recommendations intent of the treatment as defined during the prepara-
The promise of improved clinical outcomes with MRIgRT tion phase of the treatment. Therefore, procedures
comes with challenges for the treatment team, hospital need to be in place to ensure a streamlined and fast
management, and other personnel involved with the instal- decision-making process for contouring, treatment
lation and operation of MRI-linacs. Therefore, the follow- planning, and QA.
ing recommendations are separated by stakeholder groups. 9. Design dedicated documentation, available online, to
Some of the recommendations are necessarily general, for capture key aspects of the workflow for each patient
example, requiring discipline-specific specialist training, (e.g., contouring instructions, plan expectations, and
as the opportunities, availability and costs will vary between patient preparation).
62 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

9.2.2 Experienced MRIgRT Adopters 2. The emerging field of MRIgRT means that there is a
raft of technology improvements to further unlock
1. Where feasible, recruit to existing or develop new the potential of MRIgRT. Work with leading radia-
clinical trials. Ideally, some of these trials will be ran- tion oncology and radiology departments to best inte-
domized between MRIgRT and standard of care. grate these technological advances.
2. Where feasible, support existing or new clinical reg- 3. Throughput and workflow are critical to the cost-
istries (de Leon et al., 2021; de Mol van Otterloo effectiveness of MRI-linacs. Improvements in the
et al., 2021) to increase the understanding of the clin- efficiency and automation of all aspects of the work-
ical impact of MRIgRT. flow, particularly contouring, delivery time, and plan
3. Clinical benefit and health economic analyses (e.g., adaptation, are needed.
Parikh et al., 2020) should be performed to assess the 4. Support the measurement and recording of the best
upfront and ongoing costs, increase in hypofraction- estimate of delivered dose via intrafraction and inter-
ation, reduction in side effects, and quality of life, fraction dose accumulation methods (Jaffray et al.,
relative to other therapies. 2010).
4. Ongoing continuing education and a program quality 5. Support and resource efficacy and outcome-driven
review and revision are essential to maintain the clinical trials to quantify the value of MRI-linacs in
safety standards, teamwork skills, and optimal patient the health care system.
management.
5. Consider offering training courses and contributing
to education to support newer users. 9.2.5 Government
6. Consider protocolized delegation of image guidance
and adaptation tasks to appropriately credentialed 1. MRI-linacs, particularly coupled with ART, are fun-
team members with action thresholds. damentally different in the amount of information
7. Contribute to the literature for complex use cases that can be obtained and used with the intent to
and support multi-institutional benchmarking and improve patient outcomes compared with other radi-
data collection. ation therapy systems. They require additional staff
to ensure optimal use and safety. The reimbursement
for the teams and individual specialties involved in
9.2.3 Hospital Management MRIgRT should be appropriately reviewed and
1. Be aware that MRIgRT may attract patients from a revised.
wider referral market seeking advanced technologies 2. Magnetic resonance imaging–guided radiation ther-
for their cancer care. apy resource planning should be included in each
2. Prioritize MRIgRT for patients most likely to benefit government’s cancer plan.
such as upper abdominal oligometastases. 3. Invest in MRIgRT training programs to maximize the
3. While MRIgRT offers considerable advantages for appropriate use and benefit of MRIgRT.
populations expected to benefit, MRI-linacs can have 4. The emerging technology presents an opportunity for
higher capital, construction, operating, staff and edu- improvements in patient outcomes, with future
cation costs than X-ray-based linacs. Patient through- advances listed in section 9.3. Research funding
put may be reduced due to longer treatment times. should be directed toward MRIgRT technology as it
When performing site planning, these factors should has the potential to improve the scientific underpin-
be incorporated to ensure there is the financial com- nings and clinical outcomes for a large proportion of
mitment to the success and safety of the program, and all cancer patients: an excellent investment of public
the program expectations are clear to all stakeholders. funds.
4. The risk management, quality assurance, and continu-
ing education are a multidisciplinary and ongoing
process that needs to be appropriately resourced and
9.2.6 Additional Stakeholders
managed. 1. Radiation therapy professional organizations should
assess needs for MRIgRT training, additional certifi-
cation, curriculum development, and scientific con-
9.2.4 Industry
tent for their members.
1. Ensure sites and users have access to comprehensive 2. Patient advocacy groups should reach out to MRIgRT
training and offer ongoing educational activities professionals to learn more about the appropriate
through user meetings, publications, and conference patients for treatment and use their influence to
presentations. increase access for these patients.
 Summary, Recommendations, and Future Outlook 63

3. Insurance companies should carefully review the evi- tumor pathology and the MR images used for plan-
dence for MRIgRT and develop appropriate reim- ning, adapting, and targeting.
bursement policies noting the increased capital and 5. Personalize the target prescription, including a spa-
operating MRIgRT costs. Given the rapid changes in tially variable target prescription and normal tissue
MRIgRT technology and clinical findings, these poli- absorbed-dose constraints on a daily basis driven by
cies should be reviewed frequently. patient and imaging parameters, such as multipara-
metric MRI. Develop methods of biologically driven
dose delivery (Bentzen et al., 2012; ICRU Report 96
9.3 Future Outlook
Sgouros et al., 2021), responsive to changes in tumor
Magnetic resonance imaging–guided radiation therapy cre- and/or normal tissues during treatment such as those
ates opportunities for changes in the allocation of health care seen on multiparametric MRI.
resources and technical advances to improve patient 6. Link all diagnostic, treatment and follow-up informa-
outcomes. tion in registries for collective learning that are find-
able, accessible, interoperable, and reusable.
9.3.1 Health Care Resource Changes 7. Use the database information in an ongoing feedback
loop to drive contouring guidelines, prescription
1. Treat more cancers with fewer sessions (fractions) doses, and normal tissue toxicity estimates.
per treatment course, made possible by increased
accuracy and precision of absorbed-dose delivery.
This change would improve the patient experience 9.3.3 Technical Advances Needed to Further
and potentially reduce health care costs.
2. For appropriately selected patients, replace the tradi-
Improve Patient Outcomes
tional three-step image-plan-treat workflow in radio- 1. Advance technology to enable routinely treating
therapy with a single treatment session further appropriate cancer sites with real-time adaptive radi-
improving patient convenience and costs. ation therapy.
3. Perform rigorous health economic analyses to define 2. Record the absorbed-dose distribution delivered to
the cost-effectiveness of MRIgRT, taking into the tumor and normal tissue during every fraction for
account the health and economic costs of reduced every patient treated. Use this information to develop
side effects and local recurrences, where these are better and more predictive OAR constraints, facilitat-
hypothesized or proven. ing more personalized care.
4. Integrate the role of MRIgRT into clinical guidelines, 3. Develop increasingly automated, efficient, safe and
such as those provided by the (US) National robust imaging, therapy, and workflow processes.
Comprehensive Cancer Network. 4. Develop robust auto-contouring for targets and OARs
5. Assess the roles of each of the medical physicist, which is indistinguishable from clinician-derived
radiation oncologist, and radiation therapist in the contours.
MRIgRT workflow and the roles of additional health 5. Develop reliable MRI biomarkers for radiation-
care professionals from other disciplines as new indi- induced tumor and normal tissues responses that
cations emerge. develop during the course of treatment and correlate
these with absorbed dose (and equieffective dose)
and outcome measures.
9.3.2 Clinical Opportunities 6. Determine the optimal system configurations, such
1. Complete well-designed prospective randomized tri- as the magnetic field strength, imaging components,
als to provide level I evidence of the role of MRIgRT beam energy, and so on that will be appropriate for
in specific disease sites and indications. different disease sites.
2. Explore hypofractionation of radiation therapy for 7. Continue to improve the spatial and temporal capa-
tumor types that require augmented image guidance bilities of MRI anatomic and functional imaging.
and adaptation. 8. Enable adaptation during beam-on to constantly opti-
3. Investigate new oncologic and nononcologic applica- mize absorbed-dose delivery in the presence of a tar-
tions of MRIgRT to improve health outcomes. get or OAR motion or deformation, aiming for
4. Quantify appropriate margins around the target, and “margin-less” delivery.
quality the spatial variations between the underlying 9. Investigate MRI-guided proton and particle therapy.
64 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

10 Acknowledgments

The authors gratefully acknowledge the tremendous work significantly to the clinical examples. Vincent Gregoire and
of Helen Ball who contributed to many aspects of the man- Rock Mackie worked with David Rogers in the ICRU steer-
uscript preparation, revision, editing, and writing group ing committee to give guidance and feedback throughout
organization. They also thank Olga Green for assisting with the report development process. At the ICRU review meet-
framing of the report structure and many discussions related ing in Pavia in April 2022, detailed comments were received
to the report. David Rogers contributed to both the scien- from the steering committee members and María-Ester
tific writing and in an advisory role. Helen McNair contrib- Brandan, Roger Howell, and Thomas Otto. The revised
uted to the immobilization, patient setup, repositioning, report was reviewed by the steering committee members
patient comfort, and claustrophobia descriptions. The and ICRU members María-Ester Brandan, Roger Howell,
example screening forms in Appendix B were developed David Jaffray, Pawel Olko, and Norio Saito. Further exter-
and shared by the diagnostic and treatment radiographer nal reviews were received by Maria Bellon, Kevin Brown,
team at The Royal Marsden Hospital. Jinzhong Yang (C.3), John Christodouleas, Cynthia Eccles, Martin Fuss, William
Ian Zing Tan (C.4), Anna Kirby (C.4), Brigid McDonald (Bill) Hall, Aswin Hoffman, Mischa Hoogeman, Amir
(C.6), Clifton David Fuller (C.6), Suzanne Lydiard (C.7), Kishan, Luise Künzel, Bjorn Stemkens, and Esther G. C.
and Michael Christian Mayinger (C.7) wrote or contributed Troost.
 Appendix A: Table of Resources that Augment this ICRU Report 65

Appendix A: Table of Resources that

Augment this ICRU Report

Existing and under development resources that augment this ICRU report are listed in Table A.1.

Table A.1. Summary of relevant MRIgRT guideline literature to accompany this report.

Year Title and reference

2010 ICRU Report 83 on Prescribing, Recording and Reporting Photon-Beam Intensity-Modulated Radiation Therapy (IMRT) (ICRU,
2010)
2013 American College of Radiology (ACR) Guidance Document on MR Safe Practices (Kanal et al., 2013)
2014 Seminars in Radiation Oncology issue on Magnetic Resonance Imaging in Radiation Oncology https://www.sciencedirect.com/
journal/seminars-in-radiation-oncology/vol/24/issue/3
2015 MHRA Safety Guidelines for Magnetic Resonance Imaging Equipment in Clinical Use https://assets.publishing.service.gov.uk/
government/uploads/system/uploads/attachment_data/file/476931/MRI_guidance_2015_-_4-02d1.pdf
2017 Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy
Committee Task Group No. 132 (Brock et al., 2017)
2018 Clinical Oncology issue on MRI and Radiotherapy https://www.clinicaloncologyonline.net/issue/S0936-6555(18)X0012-0
2019 Seminars in Radiation Oncology issue on Adaptive Radiotherapy and Automation https://www.sciencedirect.com/journal/
seminars-in-radiation-oncology/vol/29/issue/3
2019 Clinical and Translational Radiation Oncology Journal Special issue: Online MR-Guided Radiotherapy—A new era in
radiotherapy https://www.ctro.science/radiotherapy
2020 Frontiers in Oncology issue on Magnetic Resonance Imaging in Radiation Therapy https://www.frontiersin.org/research-
topics/9123/magnetic-resonance-imaging-for-radiation-therapy
2020 AAPM Task Group 264: The safe clinical implementation of MLC tracking in radiotherapy (Keall et al., 2021)
2020 Reference dosimetry in MRI-linacs: evaluation of available protocols and data to establish a Code of Practice (de Pooter et al.,
2021)
2021 Machine QA for the Elekta Unity system: A Report from the Elekta MR-linac consortium (Roberts et al., 2021)
2021 Patterns of Care, Tolerability, and Safety of the First Cohort of Patients Treated on a Novel High-Field MR-Linac Within the
MOMENTUM Study: Initial Results from a Prospective Multi-Institutional Registry (de Mol van Otterloo et al., 2021)
2021 ESTRO-ACROP recommendations on the clinical implementation of hybrid MR-linac systems in radiation oncology (Corradini
et al., 2021)
2021 IPEM topical report: guidance on the use of MRI for external beam radiotherapy treatment planning (Speight et al., 2021a)
2021 IPEM Topical Report: an international IPEM survey of MRI use for external beam radiotherapy treatment planning (Speight
et al., 2021b)
TBD AAPM Task Group No. 351—Clinical reference dosimetry in MR-guided radiotherapy https://www.aapm.org/org/structure/
default.asp?committee_code=TG351
TBD AAPM Task Group No. 352—MR-guided Radiotherapy Systems: Considerations for clinical implementation and quality
assurance https://www.aapm.org/org/structure/default.asp?committee_code=TG352
Note. MLC = multileaf collimator; QA = quality assurance; MOMENTUM = Multi-OutcoMe EvaluatioN of radiation Therapy Using the MR-linac;
IPEM = Institute of Physics and Engineering in Medicine; MRHA = UK’s Medicines and Healthcare products Regulatory Agency; TBD = To Be
Determined since not yet approved; ESTRO-ACROP = European SocieTy for Radiotherapy and Oncology’s Advisory Committee for Radiation Oncology
Practice.
66 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Appendix B: Safety Screening Form

Examples

The example screening forms were developed and shared by the diagnostic and treatment radiographer team at The Royal
Marsden Hospital.

B.1 Initial MRI-Linac Safety Screening Questionnaire

Name: Hospital Number:

Date of birth: Height: Weight:

Please read all questions thoroughly and tick in the Yes or No boxes provided.
It is ESSENTIAL for your safety that all answers given are accurate to the best of your knowledge.
Section 1: YES NO
Do you have or have ever had a cardiac pacemaker, any pacing wires or an artificial heart valve?
Have you ever had any surgery to your brain?
Do you have any aneurysm clips?
Have you had any surgery or procedures in the last 8 weeks? (This includes any surgery, interventional procedure,
bronchoscopy, colonoscopy, endoscopy, biopsy) If so, please give details:
Have you EVER had any kind of medically implanted device placed inside your body?
If so, please give details:
Do you have a drug infusion device/syringe driver?
Have you had an injury to your eyes or body involving metal fragments or bullets?
Section 2
Do you have any artificial limbs, calipers, or supports
Do you have any braces or false teeth?
Do you wear a hearing aid?
Do you have any decorative tattoos, body piercing, or permanent make-up?
Do you have/wear any medication patches, ECG stickers, or wound dressings containing silver?
Section 3: Pregnancy Screening
Is there any chance that you could be pregnant? Date of last menstrual period:
Signature of Individual Date:
Signature of level 3 authorized person Date:
Signature of level 4 authorized person Date:

B.2 Daily MRI-Linac Safety Screening Questionnaire

Name: Hospital Number

Date of birth: Fraction:

Weight? kg YES NO
Any yes on Initial Safety Screening?
Has anything changed since last screening?
Any Medical/Dental/Surgical appointment since last screening?
Anything new on/in body since last screening?
Any metalwork undertaken since last screening?
Comments
 Appendix C: Clinical Examples 67

Appendix C: Clinical Examples

To assist new adopters of magnetic resonance imaging appendix, summarized in Table C.1. The examples are
(MRI)-guided radiation therapy with their clinical imple- ordered from, in the authors’ opinion, the ease of clinical
mentation, some clinical examples are given in this implementation.

Table C.1. Summary of the clinical examples in Appendix C.

Section Case Rationale for inclusion

C.1 Prostate Cancer Common cancer site, SABR.
C.2 Oligometastatic Progression in the Upper Respiratory motion, proximity of tumor to critical structures, difficulty
Abdomen visualizing with CBCT.
C.3 Recurrent Oligometastatic Disease in the Proximity of target to critical structures, large daily anatomic changes
Pelvis requiring adaptation.
C.4 Partial Breast Radiotherapy Visualization and margin changes, highlight skin toxicity risk due to ERE, ESE.
C.5 Multi-target Oligometastatic Case Mitigation of independent interfraction motion of multiple GTV volumes
requires online replanning.
C.6 Head and Neck Cancer Serial functional imaging during MRIgRT, adaptive RT.
C.7 Ventricular Tachycardia Combined cardiac and respiratory motion, ICD, novel treatment approach
to radiosensitive organ.
Note. SABR = stereotactic ablative radiation therapy; CBCT = cone beam computed tomography; ERE = electron return effect; ESE = electron
streaming effect; GTV = gross tumor volume; ICD = implantable cardioverter defibrillator.

C.1 Prostate Cancer Pathology, ISUP in 8/16 cores with bilateral involvement,
including 4/6 from the biopsy target lesion cores. He pre-
This case illustrates a common use case for MRIgRT with sented with a prostate-specific antigen (PSA) level of 12 ng/
ultrahypofractionation utilized for prostate cancer instead of mL and initial MRI staging of T2N0 (Fig. C.1). He received
standard or gentle hypofractionation (20 fractions). Given a consultation with a urologist and an oncologist and elected
the superior soft tissue contrast, organs at risk (OAR) as to have radical treatment with radiotherapy. He received neo-
well as the target volume are easily delineated without a adjuvant androgen deprivation therapy for 4 months prior to
separate diagnostic MRI often used for prostate radiother- radiotherapy.
apy planning. Dose delivery with ultrahypofractionation
requires higher level of precision which is enabled by
MRIgRT. There are ongoing studies to determine whether C.1.2 Treatment Intent
adaptive planning may reduce GI or GU toxicities as well as Radiation therapy is a standard treatment for localized pros-
potentially spare sexual dysfunction by sparing the neuro- tate cancer. Moderate hypofractionation (e.g., 60 Gy in 20
vascular bundle. Future directions also include further fractions) is considered standard of care in many regions.
hypofractionation with clinical trials investigating two frac- Stereotactic body radiation therapy (SBRT) is offered to
tions, dose de-escalation, and intra-prostatic boost of MRI- lower intermediate risk prostate cancer in some countries, as
defined lesions. a standard, but for ISUP Grade 3, there is less evidence sup-
porting this. This patient was recruited to the PACE C trial
C.1.1 Clinical Scenario (see Table C.2), randomizing between 20 fractions and
5-fraction SBRT, and received 5-fraction SBRT with
This is a 65-year-old man with an intermediate risk prostate MRIgRT (Brand et al., 2019). Ultrahypofractionation puts an
cancer, Gleason 4+3 (International Society of Urological enhanced onus onto the treating team, as with only 5
68 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.2. Representation of best match of rigid prostate

Figure C.1. Axial image of T2 diagnostic MRI showing 7.2-mm contours (from simulation plan; matched onto daily image with
index lesion (blue arrow) in the right peripheral zone. translations only) shown in red, onto adaptation MRI. In contrast,
the pink prostate contour is the clinician-derived contour created
on the daily adaptation MRI.

fractions any minor geometrical miss or over-dosage of

OARs may have a larger proportional effect. The prostate is
C.1.5 Organs at Risk
optimally visualized on MRI; hence, MRIgRT provides max- The OARs were contoured on the planning CT. Contoured
imal accuracy for both target and OAR. Dose intended is OARs were rectum (contoured as anorectum), bladder, non-
36.25 Gy to the planning target volume (PTV) and 40 Gy to rectum bowel, urethra, penile bulb, left and right femoral
the prostate clinical target volume (CTV; Table C.2). heads (separately).

C.1.3 Patient Positioning and Image Acquisition C.1.6 Planning Aim

Patients are asked to drink 350 mL of water 30 min prior to The planning aims are described in Table C.2. The OAR con-
simulation and treatment, to maintain a comfortably full straints applied are reproduced in Table C.3.
bladder throughout treatment. Microlette enemas were pre-
scribed for 2 days prior to the planning computed tomogra-
phy (CT) without intravenous contrast and start of C.1.7 Treatment Planning
radiotherapy, and to continue for all 5 treatment days.
Patients were immobilized using the MRI Compatible Combi Treatment planning was performed using the dedicated
Fix system (Oncology Systems Ltd, UK) and planning scans Monaco TPS version 5.40.01 (Elekta AB, Stockholm,
were acquired with a 1.5 cm slice thickness (CT) and 3 mm Sweden), using nine beams, 58 control points, and the refer-
(MRI). No MRI contrast was used. Simulation MRI was co- ence plan had 1941.2 MU.
registered with the planning CT, with the diagnostic MRI
acquired prior to pre-androgen deprivation therapy visible to
the side, to help with contouring and assessment of dominant C.1.8 Treatment Delivery and Daily Adaptation
disease location. Treatment was delivered on the Elekta Unity system (Elekta
AB, Stockholm, Sweden). Prior to each treatment fraction, a
daily adaptation T2 MRI was acquired, taking 2 min to
C.1.4 Target Volumes
acquire. The MRI sequence parameters used during planning
The target volume was the whole prostate and seminal vesi- and delivery are shown in Table C.4. The initial planning
cles deemed to be at risk of microscopic spread. contours and simulation MRI scan were transferred by the
 Appendix C: Clinical Examples 69

Table C.2. Dose objectives, as per PACE C trial (Brand et al., 2019).

Dose objective to PTV

Definition of CTV Margin for PTV (in 5 fractions) Dose objective to CTV only
Prostate +1 cm seminal vesicles 5 mm except 3 mm posteriorly 36.25 Gy isodose to cover 40 Gy isodose to cover >95% CTV
≥95% of PTV
Prostate +2 cm seminal vesicles 5 mm isotropic 30 Gy to cover ≥95% PTV None
Note. CTV = clinical target volume; PTV = planning target volume.

Table C.3. OAR constraints for 5-fraction SBRT in the PACE trial (Brand et al., 2019).

Dose-volume constraints

Maximum volume, % or cm3

Organ at risk Absorbed dose, Gy Mandatory Optimal

Rectum 18.1 50%
29 20%
36 2 cm3 1 cm3
Bladder 18.1 40%
37 10 cm3 5 cm3
a
Femoral heads 14.5 — 5%
Bowel 18.1 5 cm3
30 1 cm3
Penile bulb 29.5 — 50%
Urethra (if visualized) 42 — 50%
Note. Particular attention is paid to the rectal V36Gy and bladder V37Gy which are kept as close to 1 to 5 cm3 (respectively) as possible. SBRT = stereotactic
body radiation therapy.
a
Constraint to be achieved by each femoral head individually.

Table C.4. MRI sequence parameters used during treatment planning and delivery.

Adaptation Verification Motion management

Diagnostic T2 SFOV Simulation T2, ms image image image
Number slices 22 28 150 150 3
Field length, mm 88 70 300 300 Sag/cor/axial
Contrast T2 T2 T2 T2 bFFE
Sequence 3D TSE 3D TSE 3D TSE 3D TSE 2DsShot bFFE
Teff/TE equiv 97 110 188 188 2.0
TR, ms 5010 7210 1400 1400 4.0
Scan time, s 360 166 120 120 21
Note. TSE = turbo spin echo; bFFE = balanced fast field echo; SFOV = scan field of view; TE = echo time; TR = repetition time.

therapeutic radiographers to the daily adaptation MRI The daily adaptive plan created on the daily adaptation
through deformable image registration. Fig. C.2 shows, for image, shown in Fig. C.4, met all the required mandatory
illustration purposes, the difference between rigid registra- constraints and objectives specified in Tables C.2 and C.3.
tion, matched to prostate (red), and recontours by the physi- Prior to beam on, a further 2-min T2 MRI was acquired
cian of the day (pink). (verification MRI) and on this occasion (fraction 1) a shift
Monaco-derived deformable contours are reviewed and had occurred due to some rectal gas. A shift of the daily
edited where necessary to create the physician-approved adaptive plan (the “adapt-to-position” [ATP] workflow) was
daily contour (Fig. C.3). carried out to mitigate this change. Offline analysis showed
In general, prostate and seminal vesicle contours are that even after the shift the rectal V36Gy was out of toler-
redrawn or edited daily, but the deformed OAR contours are ance; however, re-analysis on the end-of-treatment MRI
only adjusted when significant, relevant changes have scan, taken immediately after beam off, showed that the rec-
occurred. tal V36Gy was within the optimal constraint at that point.
70 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.3. Illustration of deformable prostate contours (shown in blue) which are almost identical to clinician prostate contours
(shown in pink) (left) and those that require some editing (right). Images courtesy of Dr Kobika Sritharan.

Figure C.4. Daily adaptive plan with achieved absorbed dose-volume constraints and objectives. In the top right panel, green tick
indicates optimal and mandatory constraint met, orange exclamation mark indicates mandatory but not optimal, constraint met for this
parameter. Images courtesy of Adam Mitchell and Joan Chick, Royal Marsden Hospital. CTV = clinical target volume; PTV = planning
target volume.
 Appendix C: Clinical Examples 71

Coverage dropped during the fraction, but further analysis difficulty to visualize the lesion on standard CBCT, daily
showed that target coverage remained higher than would positional uncertainty of the target lesion, as well as its prox-
have been achieved using the simulation plan and fiducial imity to gastrointestinal structures that limit the absorbed
IGRT (image-guided radiation therapy) on a standard linac. dose and utility of ablative radiotherapy. The intent was to
use MRIgRT to provide daily personalized treatment, and
C.1.9 Treatment Outcome potentially daily adaptive replanning to optimize dose to the
target and minimize absorbed dose to the OARs.
During treatment, the patient experienced Grade 1 genitouri-
The treatment intent was to prescribe an absorbed dose of
nary side effects (mild dysuria) and no significant bowel side
40 Gy delivered in five fractions at a daily absorbed dose of
effects. By his 6-week follow-up appointment, all symptoms
8 Gy per fraction.
were back to normal. Six months after SBRT his PSA was
0.45 ng/mL with a testosterone recovered from androgen
deprivation therapy (16 nmol/L). C.2.3 Patient Positioning and Image Acquisition
The patient was positioned using a Vac-Lok (Civco Medical
C.2 Oligometastatic Progression in the Solutions, Kalona, Iowa) immobilization device with her
Upper Abdomen hands supported and stabilized above her head. Simulation
scans were performed on both a standard CT without intrave-
Stereotactic body radiation therapy in the upper abdomen nous contrast and on a Tri-60Co MRI-guided RT platform
including treatment of the pancreas, liver, abdominal lymph (ViewRay, Inc, Cleveland, Ohio). Planning MRI without
nodes has historically been challenging due to the mobile intravenous contrast was acquired using three-dimensional
nature of the target as well as that of the OARs (small bowel, (3D) steady-state free precession (TRUFI) sequence with a
large bowel, stomach). Furthermore, deformation of the OARs spatial resolution of 1.5 mm × 1.5 mm × 3 mm and a field
further complicates that ability to safely deliver ablative of view (FOV) of 45 cm × 45 cm × 24 cm. The total acquisi-
absorbed doses without harming the normal tissue. These tion time is 17 s. The simulation CT was acquired to obtain
changes are often difficult to appreciate on standard cone beam electron density information for absorbed dose calculation
CT scans. This case illustrates the use of on-table adaptive with patient in the same immobilization device using a stan-
radiotherapy enabled by MRIgRT with daily replanning to dard abdomen acquisition protocol (120 kVp, 250 mAs, 1.2
mitigate increased absorbed doses to the OARs on a daily basis mm × 1.2 mm × 1.5 mm spatial resolution). Diagnostic
and therefore reduce the risk of serious acute or chronic com- MRI and positron emission tomography-computed tomogra-
plications. Future studies are looking to see if use of on-table phy (PET-CT) were co-registered with the planning MRI and
adaptive radiotherapy can allow tumor dose escalation while CT to assist in delineating the gross tumor volume (GTV).
maintaining a low toxicity profile possibly improving disease The patient was simulated using a deep inspiratory breath-
control and potentially overall survival in these scenarios. hold approach.

C.2.1 Clinical Scenario C.2.4 Target Volumes

This is a 50-year-old woman with a history of triple-positive The target volume was the isolated enlarged peri-pancreatic
metastatic breast cancer (T3N2M1) who initially underwent lymph node as contoured on the breath-hold 0.35 T MRI
neoadjuvant chemotherapy, left-sided mastectomy, adjuvant simulation scan. This target was verified and adjusted based
radiotherapy, and was maintained on HER2-targeted thera- on the information from the CT simulation, as well as diag-
pies. Twenty months after her definitive therapy, imaging nostic MRI and PET-CT. There was no extra margin given
scan showed an isolated progressive enlargement of lymph for CTV. A PTV was created with a 5-mm isotropic
nodes between hepatic segment V/IVB and the neck of the expansion.
pancreas (Fig. C.7, left panel). She was initially considered
for radiofrequency ablation (RFA). However, this approach
C.2.5 Organs at Risk
was aborted due to the proximity of the tumor near the stom-
ach, and pancreatic duct where it was felt that there was sig- The OARs were contoured on the 0.35 T MRI simulation
nificant risk of causing ulceration using RFA. This was the scan. Adjacent OARs included stomach, small bowel, duo-
only site of disease recurrence, and thus presents to radiation denum, liver, both kidneys, and spinal cord.
oncology clinic as an oligometastatic recurrence for a discus-
sion regarding other local therapy approaches.
C.2.6 Planning Aim
The goal was to deliver 40 Gy in five fractions with 95 % of
C.2.2 Treatment Intent
the PTV covered by the prescription absorbed dose. For the
Radiation therapy, specifically SBRT, was considered for OARs, the goal was to limit the volume receiving 35 Gy (V35)
this patient. However, challenges with this case include for the stomach, duodenum, and small bowel to ≤0.5 cm3.
72 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.5. Contours, absorbed dose distribution for initial, predicted, and adapted treatment for the first fraction. Stomach (blue),
gross tumor volume (red), and planning target volume (purple) are displayed.

C.2.7 Treatment Planning Fig. C.5 shows a case of interfractional variation of the
stomach which led to online adaptive replanning for this
Treatment planning was performed using the dedicated TPS fraction. Dose-volume histograms for the initial, predicted,
(ViewRay MRIdian system) using five gantry groups, each and adapted plans are depicted in Fig. C.6.
consisting of three independent multileaf collimator modu- In Fig. C.5, note the difference in position and filling of
lated beams. the stomach for fraction 1 compared with the initial simula-
tion scan. The predicted plan shows the absorbed dose distri-
bution of the initial plan calculated on the current anatomy
C.2.8 Treatment Delivery and Daily Adaptation
resulting in excessive dose to the stomach. The adapted plan
Prior to each treatment fraction, a daily setup MRI was was re-optimized to reduce the absorbed dose to the stom-
acquired using the same acquisition protocol as the simula- ach. Isodose lines are displayed in the left box.
tion with the patient under active inhale breath-hold. The ini- Predicted plan (dotted line on the DVH) showed increased
tial planning contours were transferred from simulation MRI absorbed dose to the stomach and underdose of the target
to the daily setup MRI through deformable image registra- volumes (volume of the stomach receiving 35 Gy [V35-
tion. This was reviewed and manually adjusted by the physi- stomach] = 8.9 mL), coverage of the PTV receiving the pre-
cian and the medical physicist. scribed dose = 75 %, coverage of the GTV receiving
The initial treatment plan was then recalculated based on prescribed absorbed dose = 92.5 % compared with the initial
the patient’s current anatomy to ascertain whether the treat- plan (solid line). Adapted plan (dashed line) showed lower-
ment plan was optimal for that day’s anatomy. The decision ing of the absorbed dose to the stomach while improving
to adapt was based on target coverage (volume of PTV coverage of the target volumes (V35-stomach = 0.50 mL,
receiving the prescribed absorbed dose) and OAR constraints PTV coverage = 88 %, GTV coverage = 99.5 %).
(V35 Gy) for the stomach, small bowel, and duodenum and
whether the predicted plan on that day’s anatomy exceeded
the prespecified volumetric constraints.
C.2.9 Treatment Outcome
For example, during the daily online adaptive process, it The patient was followed closely with repeat MRI,
was observed that modifications of the volume and position PET-CT, and labs. Her first PET-CT at 3 months post MRI-
of the surrounding OARs led to a decision to adapt the plan. guided SBRT showed a complete treatment response at the
Adaptive planning was deemed necessary for all five frac- target lesion, confirmed by subsequent surveillance imag-
tions due to the mobility of the GTV as well as excessive ing (Fig. C.7). She experienced no acute toxicity due to the
absorbed dose to the stomach on a daily basis. In each of the treatment. However, she did develop some burning pain in
five adapted fractions, coverage to the PTV was improved, the epigastrium several weeks after completing therapy
and absorbed dose to the stomach was reduced. For fractions that was relieved with 20 mg daily of omeprazole. She
3 and 5, the predicted absorbed dose also indicated excessive underwent an esophagogastroduodenoscopy that showed a
absorbed dose to the duodenum. Online re-optimization with nonbleeding ulcer in the stomach for which she was pre-
the original beam angles and optimization objectives was scribed omeprazole 40 mg twice daily for 10 months
performed for each of the adapted fractions. All adapted (CTCAE [Common Terminology Criteria for Adverse
treatments were delivered with real-time MRI guidance Events] grade 2). The dose of medication was decreased as
including cine-MRI breath-hold gating with 4 frames per her discomfort improved and resolved. No other long-term
second based on the GTV and a 3-mm boundary. toxicity was noted.
 Appendix C: Clinical Examples 73

Figure C.6. Dose-volume histogram for stomach (blue lines), GTV (red lines), and PTV (purple lines). Solid lines represent initial
plan, dotted lines represent predicted plan, and dashed lines indicate adapted plan. GTV = gross tumor volume; PTV = planning target
volume.

Figure C.7. PET-CT pre-SBRT and post-SBRT with MRI guidance and online adaptive radiotherapy. Pretreatment imaging shows a
hypermetabolic tumor adjacent to segment V/IVB of the liver which was absent on the first repeat PET-CT done on 11/20/15. The
response was sustained as seen on PET-CT from 10/17/17. SBRT = stereotactic body radiation therapy.

C.3 Recurrent Oligometastatic Disease leading to uncertainty in daily absorbed dose deposition.
in the Pelvis These soft tissue structures are difficult to appreciate on
daily cone beam CT scans leading to uncertainty in the
Similar to the abdomen, certain tumors and targets are dif- absorbed dose near these viscous organs. Using daily
ficult to visualize in the presence of significant adjacent MR-guided treatments with adapt to shape (ATS), planned
organ motion and deformation. In this case, the tumor is ablative absorbed dose can be accurately delivered to the
adjacent to the rectum and near small and large bowel tumor while ensuring OARs are appropriately spared in
74 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

each fraction. This approach may allow for future absorbed

dose escalation to improve clinical outcomes.

C.3.1 Clinical Scenario

This is a 64-year-old woman, former smoker with a history
of localized breast cancer more than 16 years ago, who was
diagnosed with metastatic adenocarcinoma of the lung 3
years prior. Molecular analyses revealed KRAS G12A and
PDL1 was 0 %. At presentation, she had involvement of the
spine requiring decompressive surgery and palliative radia-
tion therapy to T5-T9 vertebrae. She was started on sys-
temic therapy with carboplatin, pemetrexed, and
pembrolizumab with progression in the brain 1.5 years Figure C.8. Axial CT of the pelvis demonstrating a left-sided
later. She received gamma knife radiosurgery. She was pre-sacral lymph node ~3 cm in size adjacent to the rectum and
switched to a new investigational systemic regimen with posterior to small bowel and rectosigmoid colon.
docetaxel and ramucirumab and has been on this therapy
for 2.5 years. A recent CT scan of the chest abdomen pelvis
showed stable disease overall except for a peri-rectal, pre- reconstructed in-slice pixel spacing of 0.73 mm. These plan-
sacral lymph node that increased in size from 2.5 to 3 cm ning images were co-registered for target and OAR delinea-
(Fig. C.8). A colonoscopy was performed which did not tion and treatment planning.
reveal a colorectal primary malignancy. A multidisciplinary
discussion ensued, and this was deemed an oligometastatic
C.3.4 Target Volumes
site of progression. Radiation therapy in the form of SABR
was recommended. The GTV was defined as the visible soft tissue tumor on T2
noncontrast 1.5 T treatment planning simulation MRI (Fig.
C.9) and adjusted to include the soft tissue component on
C.3.2 Treatment Intent the planning CT. No CTV was specifically defined. The
Radiation therapy for synchronous and metachronous oligo- PTV was a 5-mm isotropic expansion of the GTV except for
metastatic non–small-cell lung cancer is an emerging area of the anterior and right lateral aspect where a 3-mm expansion
investigation. Retrospective and prospective studies have was used due to proximity to the rectum and bowel loops.
suggested that ablative radiation therapy in patients with
limited sites of disease may translate into improved disease
control, quality of life, and survival in these patients (Gomez C.3.5 Organs at Risk
et al., 2016; Palma et al., 2019). As such, and based on a
The OARs were contoured on the simulation 1.5 T T2 non-
multidisciplinary discussion, a course of SABR to an
contrast MRI scan (Fig. C.9). Contoured OARs were rectum,
absorbed dose of 40 Gy in five fractions to the PTV, with
adjacent bowel loops (rectosigmoid colon), and left sacral
escalation of the absorbed dose to the GTV to 50 Gy in five
nerves. For the daily adaptive replanning workflow, only
fractions using a simultaneous integrated boost approach
OARs within a 3 cm volumetric ring around the PTV were
was recommended.
accurately re-contoured. Outside of this ring, there would be
little or no dosimetric impact on OAR contours or the ability
C.3.3 Patient Positioning and Image Acquisition to meet specified OAR dose limits.
The patient was set up with her arms above her head with
support. Specific instructions for bladder and rectal filling C.3.6 Treatment Planning
were not provided given the location of this target. The
patient was immobilized using an MRI Compatible Vac-Lok Treatment planning was performed using the dedicated
cushion. Simulation included a CT scan without intravenous Monaco TPS version 5.40.1. The absorbed dose was pre-
contrast on a Philips Brilliance 16-slice CT scanner followed scribed to cover at least 95 % of the PTV for 40 Gy and to
by an MRI scan without intravenous contrast on the Unity cover the GTV for 50 Gy (simultaneous integrated boost
MRI-linac. The CT scan has a slice thickness of 2.5 mm, and approach). The plan was first created based on the CT scan
in-slice FOV of 60 cm. The MRI scans included the acquisi- using the step-and-shoot intensity-modulated radiation ther-
tion of 3D T1- and T2-weighted sequences, with a slice apy (IMRT) technique with 11 beams and 79 segments, and
thickness of 2 mm, the spacing between slices of 1 mm, and the reference plan had 3019 MU. Then the CT-based
 Appendix C: Clinical Examples 75

Table C.5. Planning goals for PTV and OAR for initial plan and
daily ATS.

Structure Dosimetric criterion

PTV_pre-sacral_40 V40 Gy >95%
GTV_pre-sacral_50 V50 Gy >99%
Bowel loops V33 Gy <0.5 cm3 and Dmax <35 Gy
Sacral nerves V30 Gy <5 cm3
Dmax <32 Gy
Rectum V40 Gy <5%
V36 Gy <14%
V32 Gy <20%
V20 Gy <50%
Note. PTV = planning target volume; OAR = organs at risk; ATS = adapt
to shape; GTV = gross tumor volume.

Figure C.9. Axial 1.5 T2 noncontrast treatment planning

simulation MRI: red: GTV, aqua: PTV, green: rectum, brown:
nearby bowel loops, pink: sacral nerves (bottom). Axial
representation including a 3-cm ring structure (blue) on a
representative daily MRI prior to ATS. GTV = gross tumor
volume; PTV = planning target volume; ATS = adapt to shape.

reference plan was adapted to the T2-weighted MRI scan

acquired the same day with the CT scan using an offline ATS
workflow in the Monaco TPS. All contours on the reference
CT were deformed to the T2-weighted MRI scan, reviewed
and modified by a radiation oncologist. A synthetic CT based
on bulk density override was created based on the deformed
contours for absorbed dose calculation. The beam arrange-
ments and IMRT constraints were copied from the CT-based Figure C.10. Dose color wash for the treatment plan prior to
the first fraction (top). CT-based reference plan was generated
reference plan to the MRI-based plan for re-optimization.
followed by an offline ATS to the T2-weighted treatment planning
The MRI-based plan was used as reference for daily plan simulation MRI scan obtained the same day. Red displaying a
adaptation during treatment. The planning goals are shown minimum of 40 Gy with >88% GTV receiving at least 50 Gy and
in Table C.5. 95% of the PTV receiving at least 40 Gy; dose-volume histogram
displaying volumetric absorbed doses to the GTV, PTV, Bowel
loop, and Rectum (bottom). OAR absorbed dose constraints
C.3.7 Treatment Delivery and Daily Adaptation were all met in this plan and all subsequent ATS plans. ATS =
adapt to shape; GTV = gross tumor volume; PTV = planning
Treatment was delivered using the Elekta Unity system.
target volume; OAR = organs at risk; DVH = dose volume
Prior to each treatment fraction, a daily adaptation T2 MRI histogram.
was acquired with an acquisition time of 6 min. Similar to
the offline ATS workflow, the initial planning contours
were transferred to the daily adaptation MRI through physician as daily variation in rectal and bowel filling as
deformable image registration. The accuracy of the GTV well as positioning varied due to interfraction motion
and PTV contours was verified and edited by the treatment within the 3-cm volumetric ring structure. Then a daily ATS
physician. The OARs, in particular, the rectum and bowel plan was re-optimized to meet the initial target and OAR
loops, were edited or recontoured daily by the treating dosimetric goals (Fig. C.10). The final daily ATS plan was
76 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

renormalized accordingly to improve PTV coverage daily confirmed a small, unifocal grade 2 invasive ductal carci-
when OAR absorbed doses were below the intended limits noma measuring 12 mm without lymphovascular invasion or
that allow such normalization. associated ductal carcinoma in situ, clear surgical margins,
no nodal involvement (0/3) and ER+/PR+/HER2−. She
recovered well postoperatively and had been commenced on
C.3.8 Treatment Outcome
adjuvant letrozole without any adverse side effects. She was
The patient has had two subsequent PET-CT scans for sur- referred for adjuvant partial breast radiotherapy.
veillance which showed a complete metabolic response at
the treatment site without evidence of distant recurrence. She
is also without any evidence of treatment related acute or C.4.2 Treatment Intent
subacute complications as a result of this therapy. Adjuvant partial breast radiotherapy delivered in 26 Gy over
5 daily fractions is the standard of care for early-stage, low-
C.4 Partial Breast Radiotherapy risk ER+/HER2− breast cancer patients as per IMPORT
LOW (Coles et al., 2017), FAST-Forward (Murray Brunt
This case highlights the utilization of MRIgRT for partial et al., 2020), and The Royal College of Radiologists consen-
breast irradiation for early-stage breast cancer. With MRIgRT, sus (RCR, 2021) Although this treatment is conventionally
one distinct advantage is the superior soft tissue imaging, delivered on a conventional linac with daily CBCT guid-
enabling clearer delineation of the lumpectomy cavity and ance, delivery of this treatment on the MRI-linac enables
target for treatment during planning and during irradiation. direct visualization of the tumor bed without additional ion-
As such, this allows for smaller margins than traditional izing radiation exposure. To assess the safety and feasibility
CBCT-based partial breast treatments, and allows for the of this treatment on the MRI-linac, she was 1 of 10 patients
potential of less toxicity such as long-term tissue fibrosis and treated within the Prospective Evaluation of Radiotherapy
improved cosmesis. However, due to unique effects of Using Magnetic Resonance Image-Guided Treatment
MRIgRT such as the electron return effect (ERE) and elec- (PERMIT) trial (NCT03727698) and the Multi-OutcoMe
tron streaming effect (ESE), higher absorbed doses of radia- EvaluatioN of radiation Therapy Using the MR-linac Study
tion may be deposited at the air skin interface and needs to be (MOMENTUM) trial (NCT04075305) (de Mol van Otterloo
accounted for during treatment planning. The future use of et al., 2020).
this indication may allow further reduction in the total
absorbed dose and number of fractions enabling both patient-
centered care with the possibility of superior outcomes. C.4.3 Patient Positioning and Image Acquisition
The patient underwent a noncontrast enhanced planning CT
C.4.1 Clinical Scenario scan supine in free breathing with both arms up on an Elekta
This is a fit, 59-year-old postmenopausal woman who was WingSTEP M Low board. No MRI planning scan was per-
diagnosed with a screen-detected early-stage, low-risk left- formed in this case. An in-house bolus curtain support was
sided breast cancer. She underwent definitive surgery via left developed to shield the chin and upper arms from the ESE
wide local excision and sentinel lymph node biopsy which (Fig. C.11).

Figure C.11. Patient setup with bolus curtain support.

 Appendix C: Clinical Examples 77

Figure C.12. Pretreatment session daily MRI acquired on the MRI-linac (T1-weighted non-fat-suppressed 3D Tra) (left) with
corresponding no-contrast planning CT scan (right). GTV shown in red (encompassing surgical clips), CTV in green, and PTV in pink.
GTV = gross tumor volume; CTV = clinical target volume; PTV = planning target volume.

Table C.6. Absorbed dose constraints as per PERMIT.

C.4.4 Target Volumes
Structure Dose constraints
Target volumes are delineated as per IMPORT LOW (Coles
et al., 2017). The partial breast GTV is the surgical clip- PTV D99% ≥23.4 Gy
defined tumor bed cavity, while the CTV is the GTV with a D95% ≥24.70 Gy
minimum 15-mm isotropic expansion, excluding the under- D98% ≥24.7 Gy
lying chest wall and ribs, and 5 mm below the skin surface. D50% ≥25.74 Gy
D50% ≤26.26 Gy
The partial breast PTV is the CTV with an isotropic 10-mm
Contralateral breast Mean absorbed dose ≤1 Gy
expansion (Fig. C.12).
Ipsilateral breast V26 Gy ≤5%
Heart Mean absorbed dose ≤2 Gy
C.4.5 Organs at Risk V11 Gy ≤10%
Contralateral lung V1.6 Gy ≤10%
The OARs are also delineated as per IMPORT LOW (Coles Ipsilateral lung V11 Gy ≤15%
et al., 2017). These included the ipsilateral and contralateral
breasts and lungs, heart, and chest wall. Note. PERMIT = Prospective Evaluation of Radiotherapy Using Magnetic
Resonance Image-Guided Treatment; PTV = planning target volume.

C.4.6 Planning Aim

The aim is to deliver 26 Gy in five daily fractions with C.4.8 Treatment Delivery and Daily Adaptation
100 % of the prescription isodose covering 95 % of the par- Treatment was delivered on the Elekta Unity MRI-linac
tial breast PTV (Coles et al., 2017). Details are shown in via IMRT with 4—6 coplanar beams in a clinician-inde-
Table C.6. pendent ATP workflow. Prior to each daily fraction, a
T1-weighted pretreatment treatment session MRI (T1 3D
Tra non-fat-suppressed) was acquired, from which a daily
C.4.7 Treatment Planning ATP plan is generated prior to automated plan checking
Treatment planning was performed using Monaco TPS (Fig. C.13).
5.40.01 using a class solution. Exit doses were avoided Motion monitoring with 2D cine-MRI (btFFE Cor Sag
through the heart and contralateral breast. Multicriterial opti- Tra RealTime) was acquired throughout treatment delivery,
mization was used during optimization to drive mean contra- and a T1-weighted posttreatment session MRI (T1 3D Tra
lateral breast absorbed dose below 1 Gy. non-fat-suppressed) was also acquired.
78 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.13. Example of an ATP plan showing isodose colorwash, dose-volume histogram, and dose constraints. ATP = adapt to position.

C.4.9 Treatment Outcome (PSMA-PET) scans revealed no tumor progression; how-

ever, PSA rose from 0.4 to 3.8 μg/L over the last 3 years.
The patient completed radiotherapy with mild CTCAE grade In parallel, MRI showed increase of volume for three
1 fatigue, breast redness, tenderness, and induration, which small lymph nodes, which were evaluated to be oligome-
had improved at her last review at 6 weeks post treatment. tastases. This resulted in an indication for radiotherapy for
She will undergo surveillance at 3, 6, 12, and 24 months as these three lesions. This case is also described for the
per PERMIT and MOMENTUM protocol, in addition to rou- evaluation of the first single and multiple lymph node
tine mammographic surveillance. oligometastases on the Elekta Unity system by Winkel
et al. (2020). Fig. C.14 shows two of the three lymph
C.5 Multitarget Oligometastatic Case nodes on CT and MRI; the third node is not in this plane.
The distance between these two nodes varies over the
This case highlights a common clinical dilemma when treat- course of the treatment, as well as the orientation relative
ing multiple lesions and isocenters with a single plan and to the surrounding anatomy. In this example, the volume
how best to mitigate multiple independent interfraction of the bowel seems to impact the relative distance of these
motion. Without adaptation, the treatment team needs to two lymph nodes. Daily adaptation should account for
decide either to favor targeting one of the lesions, for exam- these changes and as such deliver the intended dose as
ple, the largest, comprising coverage of the other lesions, or defined in the pretreatment setting.
to average the shifts to achieve the best compromise of target
coverage over all lesions. However, with the ability of
MRIgRT to daily adapt the treatments based on daily loca- C.5.2 Treatment Intent
tion of the tumors, one can optimize coverage for all the tar- Stereotactic body radiation therapy for lymph node oligome-
gets while still optimally sparing normal tissue from tastases (Tree et al., 2013) is the treatment option of choice
excessive radiation. for this patient. Stereotactic body radiation therapy implies
high dose delivered in few fractions, making image guidance
C.5.1 Clinical Scenario imperative for precise targeting. In this case, the patient was
treated with curative intent on the Elekta Unity system in
This case study is from a 66-year-old man with a history five fractions of 7 Gy, with daily ATS plan optimization
of prostate cancer. Six years earlier, the patient underwent (Winkel et al., 2019) to account for interfraction deforma-
prostatectomy. In the follow-up, repeated prostate-spe- tions and motion of the three involved lymph nodes relative
cific membrane antigen positron emission tomography to each other.
 Appendix C: Clinical Examples 79

Figure C.14. (A) Pretreatment CT with the online GTV contours for each fraction. The colored contours each represent the
contours used in one of the five online fractions and correspond to those in the MRI images. (B) Pretreatment situation on MRI and (C)
and (D) online MRI and GTV and PTV contours of these two targets for fractions 1 and 2, respectively. In fraction 1, a large variation
relative to the pretreatment and to fraction 2 can be observed by evaluating the Euclidean distance between the center of gravities
of both targets. The distance decreased from 34.8 to 30.6 mm for this particular case because of the variation and the corresponding
impact in bowel (dark blue) volume and shape. GTV = gross tumor volume; PTV = planning target volume.

C.5.3 Patient Positioning and Image Acquisition guidance, a 3D T1 FFE MRI scan with an acquired voxel
size of 1.5 × 1.5 × 2.0 mm3 was acquired.
Pretreatment CT (Philips Medical Systems, Best, the
Netherlands) and multisequence MR imaging (Philips
Medical Systems) were performed to identify the involved
C.5.4 Target Volumes
lymph nodes and the nearby OARs. The CT scanning was Three lymph node metastases are identified as targets based
performed using a special table overlay to enable reproduc- on T1- and T2-weighted MRI scans. The involved lymph
ible patient setup on the MRI-linac using specific couch nodes are defined as the GTV and an isotropic 3-mm PTV
index points. In-plane resolution was 1.1 × 1.1 mm2, and the margin around the GTV was used to take geometrical uncer-
slice thickness was 2 mm. During pretreatment imaging and tainties into account.
during treatment, the patient was immobilized by using a
vacuum mattress (BlueBAG; Elekta AB, Stockholm,
Sweden).
C.5.5 Organs at Risk
The pretreatment MRI scans were registered to the CT for The nearby OARs, deliniated on T2 and T2 MRI, were the
delineation and included 3D T1, 3D T2, sagittal T1 and T2, bowel, urethra, rectum, sigmoid, the cauda equina, and the
transversal T1- and T2-weighted scans. For treatment, the left and right sacral plexus nerves. A structure including the
patient was positioned, using the couch index point, in the volume of the three PTVs plus a 2-cm margin was defined.
vacuum mattress on the MRI-linac couch. For daily image This structure is used as boundary for editing the OAR
80 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

contours; only contours within this boundary are edited as electron densities was done: bone structures were assigned
variations in dose to the OARs further away are assumed to the average bone density from the pretreatment CT; the
have negligible impact on the daily plan adaptation. other structures were assigned a relative electron density of
1. The optimize weights and shapes option was used for the
online plan optimization; beam angles and calculation grid
C.5.6 Planning Aim
size the same as for the pretreatment planning; the OAR
During the pretreatment planning, the dosimetric aim for the dose constraints were prioritized above the PTV coverage.
target lesions was to reach 35 Gy over five fractions for at For all fractions, the GTV and OAR constraints were met.
least 95 % of the PTV volume with a maximum dose 47.25 Treatment sessions took, on average, approximately 45 min.
Gy to a volume of 0.1 cm3.
For the OARs, the following constraints were used:
C.5.9 Treatment Outcome
•• For bowel not more than 0.5 cm can receive 32 Gy,
3
The patient responded well to the treatment with no toxicity
while not more than 10 cm3 can receive 25 Gy. reported. Follow-up showed an initial decrease in PSA but
•• The urethra should not receive more than 40 Gy. after 2 years PSA rise led to a new diagnosis of a tumor
•• For rectum and sigmoid, the maximum dose should recurrence at the resection margin of the prostatectomy from
not exceed 40 Gy, while not more than 0.5 cm3 should 8 years earlier. Patient has been treated for that lesion with
receive more than 32 Gy. local radiotherapy on Unity with a 5 × 7 Gy SBRT.
•• The cauda equina should not receive more than
28 Gy.
C.6 Head and Neck Cancer
•• For the sacral nerves, not more than 0.1 cm3 should
receive 32 Gy. This case illustrates the ability of the MRI-linac to address a
course of fractionated chemoradiation therapy for localized
During the online treatment planning, that is, with the patient and advanced head and neck cancer. Due to the bulky nature
on the treatment table, the same constraints were used, with of the disease as well as the expected early and good
the exception that the 10-cm3 constraint on the bowel and the response to therapy during the course, significant tumor
0.5-cm3 constraint on rectum and sigmoid were dropped. shrinkage and patient weight loss is anticipated. As such,
Basically, only the maximum dose, or proxies thereof, is without serial monitoring and adaptive therapy, there is a
used for the online treatment planning to speed up the high likelihood that tumor coverage may be compromised
process. while normal tissue would be overdosed. This case thus
nicely illustrated the workflow to monitor these changes
during daily fractions while triggering offline adaptive ther-
C.5.7 Treatment Planning apy on a weekly basis, on average, to account for these
Treatment planning was performed with the Monaco 5.40 changes. In addition, given the radiosensitive nature of the
treatment planning system (Elekta AB). A 10-beam (7 MV disease, functional imaging with techniques such as diffu-
Field Flatness Free [FFF] beam) step-and-shoot IMRT plan sion-weighted imaging (DWI) was employed to track
was generated for that fulfilled the dosimetric constraints. response to therapy. In the future, an opportunity to de-esca-
Dose calculation in the pretreatment planning was done on late radiation dose to responding subvolumes of disease and
the CT scan. This treatment plan is the reference plan for escalate radiation dose to radio-resistant subvolumes of dis-
online planning. ease can be validated through protocols that evaluate out-
The calculation grid was 2 × 2 × 2 mm3. come based on this biomarker.

C.5.8 Treatment Delivery and Adaptation C.6.1 Clinical Scenario

The treatment was delivered in five sessions on the Elekta A 73-year-old male never-smoker was seen for a base-of-
Unity system. Each session started with a daily 3D tongue mass and bilateral neck adenopathy. Evaluation
T1-weighted FFE scan with an acquired voxel size of 1.5 × revealed a base of tongue mass with adenopathy. On tran-
1.5 × 2.0 mm3. The ATS (Winkel et al., 2019) strategy was soral examination, an exophytic mass in the right mid/lower
followed to generate the daily adapted plan. For this the tonsillar fossa without soft palate or pillar invasion was visu-
contours from the pretreatment CT were propagated to the alized. Neck examination showed a palpable right upper
daily MRI using nonrigid image registration. The target and neck mass in levels II and III, 4.5 cm in aggregate.
OAR contours were manually assessed and edited with the A diagnostic CT neck showed a base of tongue lesion
ring of 2 cm around the envelop of the three PTVs. For dose mass and bilateral upper neck adenopathy, right greater than
calculation on the daily MRI, bulk assignment of the left, as indicated in Fig. C.15.
 Appendix C: Clinical Examples 81

Figure C.15. Pre-therapy diagnostic CT axial (left), coronal (center), and sagittal slices (right), demonstrating the base of tongue lesion
(blue arrow) and bilateral lymphadenopathy (yellow arrows).

Positron emission tomography-computed tomography

identified a fluorodeoxyglucose (FDG) avid mass in the right
base of tongue and oropharynx with Standard Uptake Value
(SUV) 24.3. Bilateral FDG avid cervical nodes were identi-
fied: a predominantly necrotic right level II node measuring
3.7 cm × 3.1 cm (SUV 14.8); a right level III node measur-
ing 3.1 cm × 2.7 cm (SUV 15.8), a left level II node which
is also partially necrotic measuring 2.3 cm × 2.1 cm (SUV
9.7), a left infraclavicular node measuring 1.1 cm × 0.9 cm
(SUV 7.9). An overview of tumor and nodal measurements is
shown in Fig. C.16, with extent of disease shown on PET-CT
in Fig. C.17. An ultrasound-guided nodal biopsy confirmed
human papilloma virus (HPV16) positivity. With clinical
examination and given HPV status, the patient was staged as
a cT3N2M0 (AJCC 8th ed. Summary Stage III) and was dis-
positioned to concurrent chemoradiotherapy with concurrent
cisplatin.

C.6.2 Treatment Intent

Figure C.16. Representative 18FDG PET-CT slice, showing
Adaptive MRI-based replanning was considered due to the cross-sectional measurement of disease in axial plane for bilateral
bulky exophytic nature of disease and the high likelihood of nodes and primary. FDG = fluorodeoxyglucose.
rapid tumor response and patient weight loss during therapy.
The treatment plan called for 33 daily fractions of simulta- assessment, using offline adaptive planning and online
neous integrated boost IMRT at 2.1 Gy per fraction, with adaptive positional verification on an Elekta Unity
OAR sparing of listed OARs. The patient was placed on an MRI-linac.
observational adaptive protocol, and prescribed using an
ART approach deemed “ART ex aequo” (Heukelom and Fuller,
2019), wherein each new adaptive plan has the same con-
C.6.3 Patient Positioning and Image Acquisition
straints for tumor and OAR dose as the original simulation, The patient received immobilization with S-frame, inte-
and the intent is for prescription stability rather than inten- grated intra-oral stent, and custom-mold occipital immobili-
tional GTV shrinkage, aggressive directed sparing of OARs, zation, and was simulated with noncontrast CT, 1.5 T
nor additional dose to the tumor (i.e., isotoxic, isotreatment MRI-linac hybrid device simulation, and pre/postcontrast
ART). A “serial” ART typology was used, wherein serial multiparametric MRI simulator scans. The CT acquisition
daily images were compared with the base (day 0) treatment was performed using a 120 kVp, with 2-mm slice width and
plan as a reference, without serial dose accumulation scan length from the patient’s cranial vertex to their carina.
82 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.17. CT (left), co-registered PET-CT (center), and 18FDG PET (right) in axial (top), coronal (middle), and sagittal (bottom)
planes, to illustrate complete extent of pre-therapy disease. FDG = fluorodeoxyglucose.

This was followed by 1.5 T 16-channel PA/flex coil-aided deliver 69.30 Gy in 33 fractions to high-risk disease (i.e., gross
acquisition using T2W (with and without fat suppression, tumor volume-primary [GTVp], and nodes [GTVn]), plus a
FS), fan-based reconstruction echo-planar (EPI) DWI, T2 CTV accounting for subclinical spread and a PTV constructed
map, T2* map, T1 map, 4D dynamic contrast-enhanced via a geometric 3-mm CTV-to-PTV uniform expansion. Low-
(DCE), postcontrast T1W, and variable flip angle “black risk (56 Gy) and intermediate-dose CTV/PTV volumes (60/66
bone” scans. MRI-linac simulation included a 2-min non-fat- Gy) were similarly defined to encompass elective neck and lym-
suppressed T2 sequence (T2 TRA) for positional verification phatic regions. An indicative image display of CT- and FS-T2W
daily, as well as a 6-min non-fat-suppressed T2 version for images with relevant GTVs/CTVs is shown in Fig. C.20.
OAR assessment, a SPAIR (SPectral Attenuated Inversion
Recovery) T2 for target delineation/adaptation, a turbo spin
C.6.5 Organs at Risk
echo (TSE) T2, T1 W, and EPI and TSE DWI acquisitions.
Representative axial slices from the CT and MRI contrast- The OARs were contoured on the CT simulation images, as
enhanced simulation used for dose calculation and target/ well as the 1.5 T MRI simulation with bony regions such as
OAR definition are shown in Fig. C.18, while the MRI-linac mandible, segmented via black bone MRI. Parotids were
sequences used for both initial simulation and adaptive segmented on T1W and T2W sequences, while fat-sup-
replanning/monitoring are illustrated in Fig. C.19. pressed and standard T2 sequences were used for nerves
(e.g., optic nerves, brachial plexus). Contoured OARs
included mandible, bilateral parotids and submandibular
C.6.4 Target Volumes glands, spinal cord, brainstem, brain, cochleae, optic nerves,
The planning MR and CT in immobilization were rigidly regis- lens, eyes, and brachial plexus, among others (Fig. C.21).
tered, and deformably registered to the pre-therapy staging
PET-CT. The pre-therapy target volume was constructed as visi-
C.6.6 Planning Aim
bly evident disease on fat-suppressed T2 imaging/T1 postcon-
trast, TSE DWI, and fused FDG PET-CT. Positional registration The planning goals are given in Table C.7.
and daily alignment were pre-verified using CT to T2W MRI
fusion in the treatment planning system for electron density
assignment and dose calculation. Following an in-house (MD
C.6.7 Treatment Planning
Anderson) protocol, a radiotherapy plan was constructed and Treatment planning was performed using the dedicated
approved through multiphysician quality assurance process to Monaco TPS version 5.40.1. The dose was prescribed to
 Appendix C: Clinical Examples 83

Figure C.18. Axial slice showing tumor (blue arrow) and nodal volumes (yellow arrows) across noncontrast CT simulation (top left),
fat-suppressed T2 (top right), T1 Dixon (bottom left), and variable flip angle “black bone” (bottom right) imaging for dose calculation
and target/OAR delineation pre-therapy. FS = fat suppression.

cover at least 95 % of the PTV for the prespecified volumes C.6.8 Treatment Delivery, Serial Adaptation, and
as per Table C.7. The plan was first created based on the CT Quantitative Imaging
scan using the step-and-shoot IMRT technique with 13
beams and 6 segments, and the reference plan had 90.94 The beam arrangements and IMRT constraints were copied
MU. Then the CT-based reference plan was adapted to the from the CT-based reference plan to the MRI-based plan for
T2-weighted MRI scan acquired the same day with the CT re-optimization using a daily “virtual couch shift” (ATP). The
scan using an offline ATS workflow in the Monaco TPS. MRI-based plan was used as reference for daily plan adapta-
All contours on the reference CT were rigidly (day 0) and tion during treatment. Each day, a 2-min non-fat-suppressed
deformably (day 1) registered to the T2-weighted MRI T2 TRA sequence was used for daily positional alignment
scan, reviewed and modified by a radiation oncologist. A and ATP virtual couch shift, followed by “warm-start” adap-
synthetic CT based on bulk density override was created tive re-optimization. During this process, a fat-suppressed T2
based on the deformed contours for dose calculation. An SPAIR sequence was acquired over 6 min for potential use for
indicative treatment plan from the first week of therapy is target definition via an offline “ATS” full re-optimization,
shown in Fig. C.22. followed by a monitoring EPI DWI acquisition during “beam
84 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.19. Selected MRI-linac simulation images acquired in immobilization for the same slice level and usable for serial adaptation;
note differential conspicuity and contrast/noise ratio of the primary, nodal, and OAR tissues across the same axial plane. FS = fat
suppression; SPAIR = SPectral Attenuated Inversion Recovery; TSE = turbo spin echo; EPI = echo-planar imaging; DWI = diffusion-
weighted imaging.

Figure C.20. Representative CT and fat-suppressed T2 images, showing CTVs for dose prescription, as per embedded color legend. Note that in
this case the CTV includes air cavities which are not the true definition of the CTV. CTV = clinical target volume; GTV = gross tumor volume.
 Appendix C: Clinical Examples 85

Figure C.21. Matched CT and fat-suppressed T2 images for organ-at-risk segmentation, showing OARs as per embedded color legend.

Figure C.22. Representative pre-therapy IMRT plan, showing relative isodose coverage as per embedded color legend.
86 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Table C.7. Planning goals for target volume and OAR for initial on” after RT/Physics approval of dose/constraint metrics
plan and daily ATS. (Table C.7). The patient was adapted during the first week of
Structure Dosimetric criterion therapy due to tumor volume loss to adjust for overcoverage
of parotid OARs due to tumor shrinkage and had four subse-
Spinal cord Max <40 Gy quent offline ATS replans due to weight loss to ensure target
Spinal cord PRV Max <44 Gy coverage and avoid overdosage of OARs due to positional
Brain Max <54 Gy; Goal V40 <5% alteration; Fig. C.23 illustrates reduction in tumor/node vol-
Brainstem Max <50 Gy; V30 <30% ume and patient weight loss/volume alteration between first
Mandible Max <70 Gy; V40 <40% week of therapy and last ATS during final week of treatment.
Parotids Mean <26 Gy; V15 <50%
Serial daily EPI DWI tumor monitoring was performed for
Eyes/Optic nerves Max <45 Gy; ALARA
functional imaging assessment, allowing iterative measure-
Lenses Max <5 Gy
ment of alteration in ADC during therapy; iterative fractional
Larynx/Esophagus Mean <30 Gy
ADC values are detailed in Fig. C.24.
Cochlea Max <35 Gy
Brachial plexus Max <66 Gy
CTVs V100 >95%; V95% >99; V105 <10%;
Dmax <110%; minimize V105% outside C.6.9 Treatment Outcome
GTV
PTVs Dmax <107%; Dmean <102% The patient tolerated treatment with acute grade 3 dysphagia,
oral mucositis and fatigue, and grade 2 dermatitis and grade 3
Note. Note that instead of maximum dose, ICRU 83 recommends a small pain requiring gabapentin and narcotics acutely. He was
percentage or volume be used (ICRU, 2010). ATS = adapt to shape;
PRV = Planning Organ at Risk Volume; CTV = clinical target volume;
briefly hospitalized and treated while in-patient during ther-
PTV = planning target volume. apy, requiring temporary nasogastric tube placement without

Figure C.23. Matched slice number serial axial SPAIR T2 MRI-linac acquisitions during adaptation, showing differential in tumor/
node and body volume across treatment; note significant parotid shrinkage with weight loss. SPAIR = SPectral Attenuated Inversion
Recovery.
 Appendix C: Clinical Examples 87

Figure C.24. (A) Case study depicting the change in apparent diffusion coefficient (ADC) of a primary tumor (GTV_p) and two
pathological lymph nodes on the left and right (GTV_n_L and GTV_n_R, respectively). (A) GTV_p (red), GTV_n_L (green), and
GTV_n_R (blue) contoured on T2-weighted MRIs and ADC maps acquired weekly on a 1.5 T MRI-linac. (B) Change in mean ADC of
each volume throughout radiation therapy treatment. Error bars represent the standard deviation of ADC values within the volume.
ADC = apparent diffusion coefficient; GTV = gross tumor volume.

permanent feeding tube despite significant weight loss. At 9 C.7 Ventricular Tachycardia
weeks post-therapy, contrast CT examination (Fig. C.25)
showed “Near complete interval resolution of prior base of This clinical example of treating ventricular tachycardia
tongue carcinoma with residual asymmetric soft tissue inspec- (VT) is summarized from Mayinger et al. (2020) and dem-
tion” with “Significant decrease in bilateral cervical adenopa- onstrates advanced MRIgRT capabilities. While the initial
thy with residual nodes measuring up to 2.1 cm likely experience treating VT is with traditional linacs, there are
reflective of treated nodal remnants, to be followed.” A subse- certain advantages of using an MRIgRT approach given the
quent 12-week 18FDG PET-CT (Fig. C.25) noted “Interval superior soft tissue imaging, ability for real-time tracking of
resolution of abnormal activity in the primary in the right base OARs. In this particular case, given the large volume of the
of tongue. No residual FDG avid disease is seen in the oro- target lesion, a cardiorespiratory internal target volume
pharynx and right glossopharyngeal sulcus. There is interval (ITV) approach was not feasible. Given the multidisci-
reduction in size and activity of the right cervical level 2 and plinary expertise required, and the risk involved of a high-
3 adenopathy (with SUV of 2.7 and 3.8, in keeping with post dose single fraction treatment to the heart, a radiosensitive
radiation inflammation, as low-grade residual activity in the organ surrounded by other sensitive organs, careful plan-
right neck may relate to posttreatment effect.” A confirmatory ning, risk analysis, and mitigation should occur prior to the
ultrasound-guided biopsy showed acellular debris, and the implementation of a similar clinical scenario. Future oppor-
patient is without locoregional recurrence at 6 months. tunities with MRIgRT include faster cine imaging enabling
88 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Figure C.25. Nine-week post-radiotherapy CT examination, showing complete response in primary site and resolving nodal remnant
(top). Twelve-week 18FDG PET-CT axial slices showing residual uptake in nodal remanent (SUV 2.7–4.8), demonstrated as inflammatory
nodes without disease on ultrasound-guided Fine Needle Aspiration (bottom). FDG = fluorodeoxyglucose; SUV = Standard Uptake
Value.

true cardiac-based gating as well as functional imaging noninvasive treatment alternative, have been promising for
approach to refine and define the target lesion and the patients with therapy-refractory sustained VT (Cuculich
arrhythmogenic substrate. et al., 2017; Robinson et al., 2019; van der Ree et al., 2020).
However, the most used cardiorespiratory (ITV) concept was
not possible for this patient due to the large anticipated target
C.7.1 Clinical Scenario volume size. With an ethical approval waiver, MRIgRT was
A 71-year-old male with a history of non-ischemic dilated chosen as the most suitable last resort treatment due to the
cardiomyopathy and severely reduced left ventricular ejec- capabilities of MRIgRT to image and track moving targets
tion fraction was presenting with almost daily life-threaten- real-time during treatment delivery with gated irradiation.
ing VT and electrical storm. This was despite optimized This was the first reported MRIgRT VT treatment using an
heart failure treatment, maximum-tolerated antiarrhythmic MRI-linac. The treatment intent was to prescribe a dose of 25
medication, two previous RFA procedures, and a previous Gy delivered in a single fraction.
surgical cryoablation procedure. Further RFA was not con-
sidered due to challenges caused by the location and size of
C.7.3 Patient Positioning and Image Acquisition
the arrhythmogenic substrate, adhesions caused from the
previous ablations, and the uncertainty of RFA success. Heart The patient was positioned supine with his arms above his
transplantation was not feasible. head for simulation imaging and treatment delivery. A CT
simulation with iodine contrast and delayed enhancement
(120 kV, 184 mAs) was acquired for treatment planning and
C.7.2 Treatment Intent dose calculation. A 0.35 T MRI-linac simulation without
Initial clinical outcomes of cardiac radioablation, a develop- contrast agent was performed: a 3D volumetric true fast
ing single-fraction stereotactic ablative radiotherapy imaging with balanced steady-state precision (TrueFISP) T2/
 Appendix C: Clinical Examples 89

T1-weighted MRI was acquired in exhalation breath-hold in

17 s, a single-slice two-dimensional sagittal cine TrueFISP
MRI was acquired free breathing for 60 s to test cine tracking
and additionally during exhalation breath-hold for 20 s to
enable the tracking algorithm to identify the target and verify
exhalation reproducibility.
Further anatomical and electrophysiology data were
collected to aid target delineation. This included a 1.5 T
diagnostic contrast-enhanced MRI, an invasive electro-
physiological study, surface electrocardiogram during
clinical VT, and invasive electroanatomical mapping from
previous ablation procedures.

C.7.4 Target Volumes

The CT and 3D MRI were fused in the treatment planning
system using a nonrigid registration. A cardiac electrophysi-
ologist, radiation oncologist, and cardiologist delineated the Figure C.26. Cardiac MRI short-axis view with the red
VT arrhythmogenic substrate, illustrated in Fig. C.26, to arrow indicating the arrhythmia focus targeted in the
define the target volume. A 2-mm margin was added to the treatment, acquired on a diagnostic 1.5 T Philips scanner with a
target volume in the anterior-posterior and lateral directions, 2D VIAB late gadolinium acquisition. VIAB is a two-dimensional
viability sequence acquired data using a T1-weighted gradient
and a 3-mm margin was added in the craniocaudal direction echo.
to define a PTV of 115.1 cm3.

a 6 MV flattening filter-free beam. The treatment plan con-

C.7.5 Organs at Risk sisted of 11 static step-and-shoot IMRT beams of 42 seg-
Delineated OARs included the lungs, esophagus, stomach, ments amounting to 9784 MU.
and coronary arteries.
The MRI and dosimetric safety of the implantable cardio-
C.7.8 Treatment Delivery
verter defibrillator (ICD) were also carefully considered. It
was ensured that ICD and ICD electrode planned doses were Treatment was delivered on a 0.35 T ViewRay MRIdian
less than 500 mGy. The MRI-suitable specifications from the MRI-linac. Initial 3D alignment of the patient was performed
ICD manufacturer dictated maximum spatial gradients of 20 using a 3D TrueFISP MRI acquired in 17 s at exhalation
T/m, maximum gradient slew rate performance per axis of breath-hold. The target volume was matched, and the PTV,
200 T/(m s), whole body averaged specific absorption rate of heart, and lung contours were verified against images in the
2.0 W/kg, and a magnetic strength of 1.5 T or 3 T. All condi- treatment plan.
tions were met except for the magnetic field strength. The Two-dimensional sagittal cine TrueFISP images were
risk to the device at 0.35 T was assumed to be similar or less acquired at a rate of 4 frames per second for cine tracking.
than that at 1.5 T. Initially, the esophagus was chosen as the tracking structure
based on prior experience gained from an MRI simulation
imaging study on a healthy volunteer. However, prior to
C.7.6 Planning Aim irradiation, this produced low tracking correlation in the
The goal was to deliver 25 Gy prescribed to the 85 % isodose patient due to ICD ring artifacts, as seen in Fig. C.27.
in a single delivery. Planned D98% was 25.6 Gy, D2% was 28.6 Instead, the liver dome was selected as the tracking struc-
Gy, and Dmedian was 27.5 Gy. The organ-at-risk constraints ture. The gating boundary was set to 5 mm and the percent-
were based on those for the single-fraction arm of the NRG age of the deformable target contour that could lay outside
Radiation Therapy Oncology Group 0915 protocol (Videtic the static gating boundary was set to 20 %, triggering auto-
et al., 2015). The final planned dose distribution is shown in matic beam holds. Direct tracking of the heart or heart sub-
Fig. C.27. structures was deemed not possible at the time. This
conclusion was based on the pretreatment MRI simulation
imaging study of the healthy volunteer where low tracking
C.7.7 Treatment Planning image correlation was observed, and from the porcine
Treatment planning was performed using the ViewRay model and phantom images where artifacts due to the pace-
MRIdian MRI-linac dedicated treatment planning system for maker were observed.
90 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

The total duration of the procedure was 2 h and 28 min.

This comprised patient setup (24 min), initial image acquisi-
tion and target localization (6 min), MRI-guided cine track-
ing (46 min), fluoroscopy restarts due to tracking
interruptions, positioning checks, and a patient toilet break.
The 46 min of MRI-guided cine tracking consisted of 24 min
beam on time. The beam on time consisted of expiration
breath-holds for 14 min (≥10 s, 55 breath-holds) and gating
in expiration phase during free breathing for 10 min (<10 s,
257 breaths).
The arrhythmia detection function of the ICD was turned
off during simulation and treatment delivery. The ICD func-
tion checks and programming were performed before and
after the intervention by a cardiac electrophysiologist.

C.7.9 Treatment Outcome

Twenty-four hours posttreatment the patient developed pro-
longed electrical storm due to recurrent VT. Drug interven-
tion was provided 48 h after treatment, which was then
tapered over 1 week. During the 3 months following, no
episodes of sustained VT or further complications associ-
ated with the treatment were reported or detected. After
cessation of ventricular arrhythmias, the patient experi-
Figure C.27. Axial cross section of the acquired simulation MRI enced rapid symptomatic improvement and augmented
in the treatment position (top) showing ring artifacts caused by quality of life.
the ICD, and the same cross section with the final planned dose
distribution (bottom) from the MRI-linac treatment planning
system. The dose range displayed in the colorwash is from 25 to
29.4 Gy. ICD = implantable cardioverter defibrillator.
 References 91

References

Acharya, S., Fischer-Valuck, B.W., Kashani, R., et al. (2016). Berlangieri, A., Elliott, S., Wasiak, J., Chao, M., and Foroudi, F.
“Online magnetic resonance image guided adaptive radiation (2022). “Use of magnetic resonance image-guided radiother-
therapy: First clinical applications,” Int. J. Radiat. Oncol. Biol. apy for breast cancer: A scoping review,” J. Med. Radiat. Sci.
Phys. 94, 394–403. doi:10.1016/j.ijrobp.2015.10.015. 69, 122–133. doi:10.1002/jmrs.545.
ACR (2015). Magnetic Resonance Imaging Quality Control Bertelsen, A.S., Schytte, T., Møller, P.K., et al. (2019). “First clini-
Manual (American College of Radiology, Reston, VA). cal experiences with a high field 1.5 T MR linac,” Acta. Oncol.
ACR (2018). Phantom Test Guidance for Use of the Large MRI 58, 1352–1357. doi:10.1080/0284186x.2019.1627417.
Phantom (American College of Radiology, Reston, VA). Bertholet, J., Knopf, A., Eiben, B., et al. (2019). “Real-time
ACR (2019). Large Phantom Testing: MRI (Revised 12-12-19) intrafraction motion monitoring in external beam radiother-
(American College of Radiology). https://accreditationsupport. apy,” Phys. Med. Biol. 64, 15TR01. doi:10.1088/1361-6560/
acr.org/support/solutions/articles/11000061035-large-phan- ab2ba8.
tom-testing-mri-revised-12-12-19-. Bohoudi, O., Bruynzeel, A.M.E., Senan, S., et al. (2017).
ACR (2020). ACR Manual on MR Safety (American College “Fast and robust online adaptive planning in stereotactic
of Radiology). https://www.acr.org/-/media/ACR/Files/ MR-guided adaptive radiation therapy (SMART) for pancre-
Radiology-Safety/MR-Safety/Manual-on-MR-Safety.pdf. atic cancer,” Radiother. Oncol. 125, 439–444. doi:10.1016/j.
Ahmad, S.B., Sarfehnia, A., Paudel, M.R., et al. (2016). “Evaluation radonc.2017.07.028.
of a commercial MRI Linac based Monte Carlo dose calcu- Brand, D.H., Tree, A.C., Ostler, P., et al. (2019). “Intensity-
lation algorithm with GEANT4,” Med. Phys. 43, 894–907. modulated fractionated radiotherapy versus stereotactic body
doi:10.1118/1.4939808. radiotherapy for prostate cancer (PACE-B): Acute toxic-
Alexander, S.E., McNair, H.A., Oelfke, U., et al. (2022). “Prostate ity findings from an international, randomised, open-label,
volume changes during extreme and moderately hypofraction- phase 3, non-inferiority trial,” Lancet Oncol. 20, 1531–1543.
ated magnetic resonance image-guided radiotherapy,” Clin. doi:10.1016/s1470-2045(19)30569-8.
Oncol. 34, e383–e391. doi:10.1016/j.clon.2022.03.022. Brock, K.K. (2011). “Imaging and image-guided radiation ther-
Almansour, H., Afat, S., Fritz, V., et al. (2021). “Prospective image apy in liver cancer,” Semin. Radiat. Oncol. 21, 247–255.
quality and lesion assessment in the setting of MR-guided doi:10.1016/j.semradonc.2011.05.001.
radiation therapy of prostate cancer on an MR-linac at 1.5 Brock, K.K., Mutic, S., McNutt, T.R., Li, H., and Kessler, M.L.
T: A comparison to a standard 3 T MRI,” Cancers 13, 1533. (2017). “Use of image registration and fusion algorithms and
doi:10.3390/cancers13071533. techniques in radiotherapy: Report of the AAPM Radiation
Alnaghy, S.J., Causer, T., Roberts, N., et al. (2020). “High resolu- Therapy Committee Task Group No. 132,” Med. Phys. 44,
tion silicon array detector implementation in an inline MRI- e43–e76. doi:10.1002/mp.12256.
linac,” Med. Phys. 47, 1920–1929. doi:10.1002/mp.14016. Bronskill, M.J., Carson, P.L., Einstein, S., et al. (1986). Report No.
Axford, A., Dikaios, N., Roberts, D.A., Clark, C.H., and Evans, 20—Site Planning for Magnetic Resonance Imaging Systems
P.M. (2021). “An end-to-end assessment on the accuracy of (American Association of Physicists in Medicine, New York).
adaptive radiotherapy in an MR-linac,” Phys. Med. Biol. 66, Bruynzeel, A.M.E., Tetar, S.U., Oei, S.S., et al. (2019). “A pro-
055021. doi:10.1088/1361-6560/abe053. spective single-arm phase 2 study of stereotactic magnetic res-
Bahig, H., Yuan, Y., Mohamed, A.S.R., et al. (2018). “Magnetic onance guided adaptive radiation therapy for prostate cancer:
resonance-based response assessment and dose adaptation Early toxicity results,” Int. J. Radiat. Oncol. Biol. Phys. 105,
in human papilloma virus positive tumors of the oropharynx 1086–1094. doi:10.1016/j.ijrobp.2019.08.007.
treated with radiotherapy (MR-ADAPTOR): An R-IDEAL Byrne, H., Le Duc, G., Lux, F., et al. (2020). “Enhanced MRI-
stage 2a-2b/Bayesian phase II trial,” Clin. Transl. Radiat. guided radiotherapy with gadolinium-based nanoparticles:
Oncol. 13, 19–23. doi:10.1016/j.ctro.2018.08.003. Preclinical evaluation with an MRI-linac,” Cancer Nano. 11,
Baldwin, L.N., Wachowicz, K., Thomas, S.D., Rivest, R., and 1–14. doi:10.1186/s12645-020-00065-5.
Fallone, B.G. (2007). “Characterization, prediction, and cor- Cai, B., Green, O.L., Kashani, R., et al. (2018). “A practical imple-
rection of geometric distortion in 3 T MR images,” Med. Phys. mentation of physics quality assurance for photon adaptive
34, 388–399. doi:10.1118/1.2402331. radiotherapy,” Z. Med. Phys. 28, 211–223. doi:10.1016/j.
Bentzen, S.M., Dörr, W., Gahbauer, R., et al. (2012). “Bioeffect zemedi.2018.02.002.
modeling and equieffective dose concepts in radiation oncol- Carbucicchio, C., Andreini, D., Piperno, G., et al. (2021).
ogy–terminology, quantities and units,” Radiother. Oncol. 105, “Stereotactic radioablation for the treatment of ventricu-
266–268. doi:10.1016/j.radonc.2012.10.006. lar tachycardia: Preliminary data and insights from the
92 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

STRA-MI-VT phase Ib/II study,” J. Interv. Card. Electrophysiol. registry for the evidence-based introduction of MR-guided
62, 427–439. doi:10.1007/s10840-021-01060-5. adaptive therapy,” Front. Oncol. 10, 1328. doi:10.3389/
Chen, L., Price, R.A., Jr., Nguyen, T.B., et al. (2004). “Dosimetric fonc.2020.01328.
evaluation of MRI-based treatment planning for prostate de Mol van Otterloo, S.R., Christodouleas, J.P., Blezer, E.L., et al.
cancer,” Phys. Med. Biol. 49, 5157–5170. doi:10.1088/0031- (2021). “Patterns of care, tolerability, and safety of the first
9155/49/22/010. cohort of patients treated on a novel high-field MR-linac within
Chen, X., Ahunbay, E., Paulson, E.S., Chen, G., and Li, X.A. the MOMENTUM study: Initial results from a prospective
(2020). “A daily end-to-end quality assurance workflow for multi-institutional registry,” Int. J. Radiat. Oncol. Biol. Phys.
MR-guided online adaptive radiation therapy on MR-linac,” J. 111, 867–875. doi:10.1016/j.ijrobp.2021.07.003.
Appl. Clin. Med. Phys. 21, 205–212. doi:10.1002/acm2.12786. de Pooter, J., Billas, I., de Prez, L., et al. (2021). “Reference dosim-
Chuong, M., Clark, M., Henke, L., et al. (2021). “Patterns of utiliza- etry in MRI-linacs: Evaluation of available protocols and data
tion and clinical adoption of 0.35 MR-guided radiation therapy to establish a code of practice,” Phys. Med. Biol. 66, 05TR02.
in the United States—Understanding the transition to adaptive, doi:10.1088/1361-6560/ab9efe.
ultra-hypofractionated treatments,” Int. J. Radiat. Oncol. Biol. de Vries, J.H.W., Seravalli, E., Houweling, A.C., et al. (2018).
Phys. 111, e510. doi:10.1016/j.ijrobp.2021.07.1400. “Characterization of a prototype MR-compatible Delta4 QA
Coles, C.E., Griffin, C.L., Kirby, A.M., et al. (2017). “Partial-breast system in a 1.5 tesla MR-linac,” Phys. Med. Biol. 63, 02NT02.
radiotherapy after breast conservation surgery for patients with doi:10.1088/1361-6560/aa9d26.
early breast cancer (UK IMPORT LOW trial): 5-year results Dearnaley, D., Syndikus, I., Mossop, H., et al. (2016). “Conventional
from a multicentre, randomised, controlled, phase 3, non- versus hypofractionated high-dose intensity-modulated
inferiority trial,” Lancet 390, 1048–1060. doi:10.1016/s0140- radiotherapy for prostate cancer: 5-year outcomes of the ran-
6736(17)31145-5. domised, non-inferiority, phase 3 CHHiP trial,” Lancet Oncol.
Constantin, D.E., Fahrig, R., and Keall, P.J. (2011). “A study of 17, 1047–1060. doi:10.1016/s1470-2045(16)30102-4.
the effect of in-line and perpendicular magnetic fields on beam Desai, V., Bayouth, J., Smilowitz, J., and Yadav, P. (2021). “A
characteristics of electron guns in medical linear accelerators,” clinical validation of the MR-compatible Delta(4) QA system
Med. Phys. 38, 4174–4185. doi:10.1118/1.3600695. in a 0.35 tesla MR linear accelerator,” J. Appl. Clin. Med. Phys.
Corradini, S., Alongi, F., Andratschke, N., et al. (2021). “ESTRO- 22, 82–91. doi:10.1002/acm2.13216.
ACROP recommendations on the clinical implementation of Ding, G.X., Alaei, P., Curran, B., et al. (2018). “Image guidance
hybrid MR-linac systems in radiation oncology,” Radiother. doses delivered during radiotherapy: Quantification, manage-
Oncol. 159, 146–154. doi:10.1016/j.radonc.2021.03.025. ment, and reduction: Report of the AAPM Therapy Physics
Cuccia, F., Alongi, F., Belka, C., et al. (2021a). “Patient positioning Committee Task Group 180,” Med. Phys. 45, e84–e99.
and immobilization procedures for hybrid MR-linac systems,” doi:10.1002/mp.12824.
Radiat. Oncol. 16, 183. doi:10.1186/s13014-021-01910-6. Dunlop, A., Mitchell, A., Tree, A., et al. (2020). “Daily adaptive
Cuccia, F., Rigo, M., Gurrera, D., et al. (2021b). “Mitigation on radiotherapy for patients with prostate cancer using a high field
bowel loops daily variations by 1.5 T MR-guided daily-adap- MR-linac: Initial clinical experiences and assessment of deliv-
tive SBRT for abdomino-pelvic lymph-nodal oligometastases,” ered doses compared to a C-arm linac,” Clin. Transl. Radiat.
J. Cancer Res. Clin. Oncol. 147, 3269–3277. doi:10.1007/ Oncol. 23, 35–42. doi:10.1016/j.ctro.2020.04.011.
s00432-021-03739-8. Eccles, C.L., Adair Smith, G., Bower L., et al. (2019). “Magnetic
Cuculich, P.S., Schill, M.R., Kashani, R., et al. (2017). “Noninvasive resonance imaging sequence evaluation of an MR Linac sys-
cardiac radiation for ablation of ventricular tachycardia,” N. tem; early clinical experience”, Tech. Innov. Patient Support
Engl. J. Med. 377, 2325–2336. doi:10.1056/NEJMoa1613773. Radn Oncol. 12, 56–63.
Cunningham, J.M., Barberi, E.A., Miller, J., Kim, J.P., and Glide- Eijkelenkamp, H., Boekhoff, M.R., Verweij, M.E., et al. (2021).
Hurst, C.K. (2019). “Development and evaluation of a novel “Planning target volume margin assessment for online adap-
MR-compatible pelvic end-to-end phantom,” J. Appl. Clin. tive MR-guided dose-escalation in rectal cancer on a 1.5 T
Med. Phys. 20, 265–275. doi:10.1002/acm2.12455. MR-linac,” Radiother. Oncol. 162, 150–155. doi:10.1016/j.
Cusumano, D., Lenkowicz, J., Votta, C., et al. (2020). “A deep radonc.2021.07.011.
learning approach to generate synthetic CT in low field Elter, A., Dorsch, S., Mann, P., et al. (2019). “End-to-end test of
MR-guided adaptive radiotherapy for abdominal and pel- an online adaptive treatment procedure in MR-guided radio-
vic cases,” Radiother. Oncol. 153, 205–212. doi:10.1016/j. therapy using a phantom with anthropomorphic structures,”
radonc.2020.10.018. Phys. Med. Biol. 64, 225003. doi:10.1088/1361-6560/ab4d8e.
de Crevoisier, R., Bayar, M.A., Pommier, P., et al. (2018). Emami, H., Dong, M., Nejad-Davarani, S.P., and Glide-Hurst,
“Daily versus weekly prostate cancer image guided radia- C.K. (2018). “Generating synthetic CTs from magnetic reso-
tion therapy: Phase 3 multicenter randomized trial,” Int. J. nance images using generative adversarial networks,” Med.
Radiat. Oncol. Biol. Phys. 102, 1420–1429. doi:10.1016/j. Phys. 45, 3627–3636. doi:10.1002/mp.13047.
ijrobp.2018.07.2006. Fallone, B.G., Murray, B., Rathee, S., et al. (2009). “First MR
de Leon, J., Woods, A., Twentyman, T., et al. (2021). “Analysis of images obtained during megavoltage photon irradiation from a
data to advance personalised therapy with MR-linac (ADAPT- prototype integrated linac-MR system,” Med. Phys. 36, 2084–
MRL),” Clin. Transl. Radiat. Oncol. 31, 64–70. doi:10.1016/j. 2088. doi:10.1118/1.3125662.
ctro.2021.09.004. Ferris, W.S., Kissick, M.W., Bayouth, J.E., Culberson, W.S., and
de Mol van Otterloo, S.R., Christodouleas, J.P., Blezer, E.L.A, Smilowitz, J.B. (2020). “Evaluation of radixact motion syn-
et al. (2020). “The MOMENTUM study: An international chrony for 3D respiratory motion: Modeling accuracy and
 References 93

dosimetric fidelity,” J. Appl. Clin. Med. Phys. 21, 96–106. Green, O.L., Henke, L.E., and Hugo, G.D. (2019). “Practical
doi:10.1002/acm2.12978. clinical workflows for online and offline adaptive radiation
Finazzi, T., van Sörnsen, de, Koste, J.R., Palacios, M.A., et al. therapy,” Semin. Radiat. Oncol. 29, 219–227. doi:10.1016/j.
(2020). “Delivery of magnetic resonance-guided single-frac- semradonc.2019.02.004.
tion stereotactic lung radiotherapy,” Phys. Imaging Radiat. Guckenberger, M., Richter, A., Boda-Heggemann, J., and Lohr,
Oncol. 14, 17–23. doi:10.1016/j.phro.2020.05.002. F. (2012). “Motion compensation in radiotherapy,” Crit. Rev.
Fischer-Valuck, B.W., Henke, L., Green, O., et al. (2017). “Two- Biomed. Eng. 40, 187–197. doi:10.1615/critrevbiomedeng.
and-a-half-year clinical experience with the world’s first mag- v40.i3.30.
netic resonance image guided radiation therapy system,” Adv. Güngör, G., Serbez, İ., Temur, B., et al. (2021). “Time analysis of
Radiat. Oncol. 2, 485–493. doi:10.1016/j.adro.2017.05.006. online adaptive magnetic resonance-guided radiation therapy
Fraass, B., Doppke, K., Hunt, M., et al. (1998). “American workflow according to anatomical sites,” Pract. Radiat. Oncol.
Association of Physicists in Medicine Radiation Therapy 11, e11–e21. doi:10.1016/j.prro.2020.07.003.
Committee Task Group 53: Quality assurance for clinical Gupta, A., Dunlop, A., Mitchell, A., et al. (2022). “Online adap-
radiotherapy treatment planning,” Med. Phys. 25, 1773–1829. tive radiotherapy for head and neck cancers on the MR
doi:10.1118/1.598373. linear Accelerator: Introducing a novel modified Adapt-to-
Fu, Y., Mazur, T.R., Wu, X., et al. (2018). “A novel MRI segmen- Shape approach,” Clin. Transl. Radiat. Oncol. 32, 48–51.
tation method using CNN-based correction network for MRI- doi:10.1016/j.ctro.2021.11.001.
guided adaptive radiotherapy,” Med. Phys. 45, 5129–5137. Hackett, S.L., van Asselen, B., Wolthaus, J.W., et al. (2016).
doi:10.1002/mp.13221. “Consequences of air around an ionization chamber: Are exist-
Gach, H.M., Green, O.L., Mackey, S.L., et al. (2022). ing solid phantoms suitable for reference dosimetry on an
“Implementation of magnetic resonance safety program for MR-linac?” Med. Phys. 43, 3961. doi:10.1118/1.4952727.
radiation oncology,” Pract. Radiat. Oncol. 12, e49–e55. Hackett, S.L., van Asselen, B., Wolthaus, J.W.H., et al. (2018).
doi:10.1016/j.prro.2021.08.008. “Spiraling contaminant electrons increase doses to sur-
Gao, Y., Yoon, S., Savjani, R., et al. (2021). “Comparison and eval- faces outside the photon beam of an MRI-linac with a per-
uation of distortion correction techniques on an MR-guided pendicular magnetic field,” Phys. Med. Biol. 63, 095001.
radiotherapy system,” Med. Phys. 48, 691–702. doi:10.1002/ doi:10.1088/1361-6560/aaba8f.
mp.14634. Hales, R.B., Rodgers, J., Whiteside, L., et al. (2020). “Therapeutic
Gao, Y., Zhou, Z., Han, F., et al. (2018). “Accelerated 3D bSSFP radiographers at the helm: Moving towards radiographer-led
imaging for treatment planning on an MRI-guided radiotherapy MR-guided radiotherapy,” J. Med. Imaging. Radiat. Sci. 51,
system,” Med. Phys. 45, 2595–2602. doi:10.1002/mp.12924. 364–372. doi:10.1016/j.jmir.2020.05.001.
Garcia Schüler, H.I., Pavic, M., Mayinger, M., et al. (2021). Hall, W.A., Kishan, A.U., Hall, E., et al. (2022). “Adaptive mag-
“Operating procedures, risk management and challenges dur- netic resonance image guided radiotherapy for intact prostate
ing implementation of adaptive and non-adaptive MR-guided cancer: How to optimally test an emerging technology,” Front.
radiotherapy: 1-year single-center experience,” Radiat. Oncol. Oncol. 12, 962897. doi:10.3389/fonc.2022.962897.
16, 217. doi:10.1186/s13014-021-01945-9. Hara, W., Soltys, S.G., and Gibbs, I.C. (2007). “CyberKnife robotic
Gassenmaier, S., Küstner, T., Nickel, D., et al. (2021). “Deep learn- radiosurgery system for tumor treatment,” Expert Rev. Anticancer
ing applications in magnetic resonance imaging: Has the future Ther. 7, 1507–1515. doi:10.1586/14737140.7.11.1507.
become present?” Diagnostics 11, 2181. doi:10.3390/diagnos- Hehakaya, C., Van der Voort van Zyp, J., Lagendijk, J.J., et al.
tics11122181. (2020). “Problems and promises of introducing the magnetic
Glide-Hurst, C.K., Lee, P., Yock, A.D., et al. (2021a). “Adaptive resonance imaging linear accelerator into routine care: The
radiation therapy (ART) strategies and technical consider- case of prostate cancer,” Front. Oncol. 10, 1741. doi:10.3389/
ations: A state of the ART review from NRG Oncology,” Int. fonc.2020.01741.
J. Radiat. Oncol. Biol. Phys. 109, 1054–1075. doi:10.1016/j. Henke, L., Kashani, R., Robinson, C., et al. (2018). “Phase I trial
ijrobp.2020.10.021. of stereotactic MR-guided online adaptive radiation therapy
Glide-Hurst, C.K., Paulson, E.S., McGee, K., et al. (2021b). “Task (SMART) for the treatment of oligometastatic or unresectable
Group 284 Report: Magnetic resonance imaging simulation in primary malignancies of the abdomen,” Radiother. Oncol. 126,
radiotherapy: Considerations for clinical implementation, opti- 519–526. doi:10.1016/j.radonc.2017.11.032.
mization, and quality assurance,” Med. Phys. 48, e636–e670. Heukelom, J., and Fuller, C.D. (2019). “Head and neck cancer
doi:10.1002/mp.14695. Adaptive Radiation Therapy (ART): Conceptual consider-
Godoy Scripes, P., Subashi, E., Burleson, S., et al. (2020). “Impact ations for the informed clinician,” Semin. Radiat. Oncol. 29,
of varying air cavity on planning dosimetry for rectum patients 258–273. doi:10.1016/j.semradonc.2019.02.008.
treated on a 1.5 T hybrid MR-linac system,” J. Appl. Clin. Med. Hewson, E., Nguyen, D.T., Booth, J.T., Hugo, G., and Keall, P.J.
Phys. 21, 144–152. doi:10.1002/acm2.12903. (2020). “Adaptive radiation therapy,” in Modern Technology
Gomez, D.R., Blumenschein, G.R., Jr., Lee, J.J., et al. (2016). “Local of Radiation Oncology, van Dyk, J., Ed., Vol. 4. 259–299.
consolidative therapy versus maintenance therapy or observa- (Medical Physics Publishing, Madison, WI).
tion for patients with oligometastatic non-small-cell lung can- Hickey, B.E., Lehman, M., Francis, D.P., and See, A.M. (2016).
cer without progression after first-line systemic therapy: A “Partial breast irradiation for early breast cancer,” Cochrane
multicentre, randomised, controlled, phase 2 study,” Lancet Database Syst. Rev. 7, Cd007077. doi:10.1002/14651858.
Oncol. 17, 1672–1682. doi:10.1016/s1470-2045(16)30532-0. CD007077.pub3.
94 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Hiraoka, M., Matsuo, Y., Sawada, A., et al. (2012). “Realization Keall, P.J., Brighi, C., Glide-Hurst, C., et al. (2022). “Integrated
of dynamic tumor tracking irradiation with real-time monitor- MRI-guided radiotherapy—Opportunities and challenges,”
ing in lung tumor patients using a gimbaled X-ray head radia- Nat. Rev. Clin. Oncol. 19, 458–470. doi:10.1038/s41571-022-
tion therapy equipment,” Int. J. Radiat. Oncol. Biol. Phys. 84, 00631-3.
S560–S561. doi:10.1016/j.ijrobp.2012.07.1493. Keall, P.J., Mageras, G.S., Balter, J.M., et al. (2006). “The man-
Hissoiny, S., Raaijmakers, A.J., Ozell, B., Després, P., and agement of respiratory motion in radiation oncology report
Raaymakers, B.W. (2011). “Fast dose calculation in mag- of AAPM Task Group 76,” Med. Phys. 33, 3874–3900.
netic fields with GPUMCD,” Phys. Med. Biol. 56, 5119–5129. doi:10.1118/1.2349696.
doi:10.1088/0031-9155/56/16/003. Keall, P.J., Nguyen, D.T., O’Brien, R., et al. (2018a). “The first
Hoffmans, D., Niebuhr, N., Bohoudi, O., Pfaffenberger, A., and clinical implementation of real-time image-guided adaptive
Palacios, M. (2020). “An end-to-end test for MR-guided radiotherapy using a standard linear accelerator,” Radiother.
online adaptive radiotherapy,” Phys. Med. Biol. 65, 125012. Oncol. 127, 6–11. doi:10.1016/j.radonc.2018.01.001.
doi:10.1088/1361-6560/ab8955. Keall, P.J., Nguyen, D.T., O’Brien, R., et al. (2018b). “Review of
Houweling, A.C., de Vries, J.H., Wolthaus, J., et al. (2016). real-time 3-dimensional image guided radiation therapy on
“Performance of a cylindrical diode array for use in a 1.5 T standard-equipped cancer radiation therapy systems: Are we
MR-linac,” Phys. Med. Biol. 61, N80–N89. doi:10.1088/0031- at the tipping point for the era of real-time radiation therapy?”
9155/61/3/n80. Int. J. Radiat. Oncol. Biol. Phys. 102, 922–931. doi:10.1016/j.
Huq, M.S., Fraass, B.A., Dunscombe, P.B., et al. (2016). “The ijrobp.2018.04.016.
report of Task Group 100 of the AAPM: Application of risk Keall, P.J., Sawant, A., Berbeco, R.I., et al. (2021). “AAPM Task
analysis methods to radiation therapy quality management,” Group 264: The safe clinical implementation of MLC track-
Med. Phys. 43, 4209–4262. doi:10.1118/1.4947547. ing in radiotherapy,” Med. Phys. 48, e44–e64. doi:10.1002/
IAEA (2001). Absorbed Dose Determination in External Beam mp.14625.
Radiotherapy (International Atomic Energy Agency, Vienna, Keesman, R., van der Bijl, E., Janssen, T.M., et al. (2020). “Clinical
Austria). workflow for treating patients with a metallic hip prosthesis
Iakovenko, V., Keller, B., Sahgal, A., and Sarfehnia, A. (2020). using magnetic resonance imaging-guided radiotherapy,”
“Experimental measurement of ionization chamber angular Phys. Imaging Radiat. Oncol. 15, 85–90. doi:10.1016/j.
response and associated magnetic field correction factors in phro.2020.07.010.
MR-linac,” Med. Phys. 47, 1940–1948. doi:10.1002/mp.14025. Kennedy, W.R., Thomas, M.A., Stanley, J.A., et al. (2020).
International Commission on Radiation Units and Measurements “Single-institution phase 1/2 prospective clinical trial of
(ICRU) (2010). “Prescribing, recording, and reporting photon- single-fraction, high-gradient adjuvant partial-breast irra-
beam intensity-modulated radiation therapy (IMRT): Report diation for hormone sensitive stage 0-I breast cancer,” Int.
83,” J. ICRU 10, 1–92. doi:10.1093/jicru/ndq002. J. Radiat. Oncol. Biol. Phys. 107, 344–352. doi:10.1016/j.
Intven, M.P.W., de Mol van Otterloo, S.R., Mook, S., et al. (2021). ijrobp.2020.02.021.
“Online adaptive MR-guided radiotherapy for rectal cancer; Kerkmeijer, L.G.W., Groen, V.H., Pos, F.J., et al. (2021). “Focal
Feasibility of the workflow on a 1.5T MR-linac: Clinical boost to the intraprostatic tumor in external beam radiother-
implementation and initial experience,” Radiother. Oncol. 154, apy for patients with localized prostate cancer: Results from
172–178. doi:10.1016/j.radonc.2020.09.024. the FLAME randomized phase III trial,” J. Clin. Oncol. 39,
Jackson, S., Glitzner, M., Tijssen, R.H.N., and Raaymakers, B.W. 787–796. doi:10.1200/jco.20.02873.
(2019). “MRI B (0) homogeneity and geometric distortion with Kim, A., Lim-Reinders, S., Ahmad, S.B., Sahgal, A., and Keller,
continuous linac gantry rotation on an Elekta Unity MR-linac,” B.M. (2020a). “Surface and near-surface dose measurements at
Phys. Med. Biol. 64, 12NT01. doi:10.1088/1361-6560/ab231a. beam entry and exit in a 1.5 T MR-linac using optically stimu-
Jaffray, D.A. (2012). “Image-guided radiotherapy: From current lated luminescence dosimeters,” Phys. Med. Biol. 65, 045012.
concept to future perspectives,” Nat. Rev. Clin. Oncol. 9, 688– doi:10.1088/1361-6560/ab64b6.
699. doi:10.1038/nrclinonc.2012.194. Kim, J., Garbarino, K., Schultz, L., et al. (2015). “Dosimetric evalu-
Jaffray, D.A., Lindsay, P.E., Brock, K.K., Deasy, J.O., and Tomé, ation of synthetic CT relative to bulk density assignment-based
W.A. (2010). “Accurate accumulation of dose for improved magnetic resonance-only approaches for prostate radiother-
understanding of radiation effects in normal tissue,” Int. J. apy,” Radiat. Oncol. 10, 239–239. doi:10.1186/s13014-015-
Radiat. Oncol. Biol. Phys. 76, S135–S139. doi:10.1016/j. 0549-7.
ijrobp.2009.06.093. Kim, T., Lewis, B.C., Price, A., et al. (2020b). “Direct tumor visual
Jia, X., Gu, X., Graves, Y.J., Folkerts, M., and Jiang, S.B. feedback during free breathing in 0.35T MRgRT,” J. Appl.
(2011). “GPU-based fast Monte Carlo simulation for radio- Clin. Med. Phys. 21, 241–247. doi:10.1002/acm2.13016.
therapy dose calculation,” Phys. Med. Biol. 56, 7017–7031. Kirby, M.C., Ryde, S., and Hall, C. (2006). Acceptance Testing and
doi:10.1088/0031-9155/56/22/002. Commissioning of Linear Accelerators (IPEM, York, England).
Kanal, E., Barkovich, A.J., Bell, C., et al. (2013). “ACR guid- Kirkby, C., Stanescu, T., Rathee, S., et al. (2008). “Patient dosim-
ance document on MR safe practices: 2013,” J. Magn. Reson. etry for hybrid MRI-radiotherapy systems,” Med. Phys. 35,
Imaging 37, 501–530. 1019–1027. doi:10.1118/1.2839104.
Keall, P.J., Barton, M., and Crozier, S. (2014). “The Australian Kishan, A.U., Lamb, J., Casado, M., et al. (2022a). “Magnetic reso-
magnetic resonance imaging-linac program,” Semin. Radiat. nance imaging-guided versus computed tomography-guided
Oncol. 24, 203–206. doi:10.1016/j.semradonc.2014.02.015. stereotactic body radiotherapy for prostate cancer (MIRAGE):
 References 95

Interim analysis of a phase III randomized trial,” J. Clin. Oncol. Latifi, K., Moros, E.G., Zhang, G., Harrison, L., and Feygelman,
40, 255–255. doi:10.1200/JCO.2022.40.6_suppl.255. V. (2019). “A method to determine the coincidence of
Kishan, A.U., Ma, T., Lamb, J., et al. (2022b). “Magnetic reso- MRI-guided linac radiation and magnetic isocenters,”
nance imaging-guided vs. computed tomography-guided ste- Technol. Cancer Res. Treat. 18, 1533033819877986.
reotactic body radiotherapy for prostate cancer (MIRAGE): doi:10.1177/1533033819877986.
Primary endpoint analysis of a phase III randomized trial,” Lee, J., Bates, M., Shepherd, E., et al. (2021). “Cardiac stereotac-
Int. J. Radiat. Oncol. Biol. Phys. 114, S92–S93. doi:10.1016/j. tic ablative radiotherapy for control of refractory ventricular
ijrobp.2022.07.507. tachycardia: Initial UK multicentre experience,” Open Heart 8,
Klawikowski, S., Tai, A., Ates, O., Ahunbay, E., and Li, X.A. e001770. doi:10.1136/openhrt-2021-001770.
(2018). “A fast 4D IMRT/VMAT planning method based Lenkowicz, J., Votta, C., Nardini, M., et al. (2022). “A deep learn-
on segment aperture morphing,” Med. Phys. 45, 1594–1602. ing approach to generate synthetic CT in low field MR-guided
doi:10.1002/mp.12778. radiotherapy for lung cases,” Radiother. Oncol. 176, 31–38.
Klein, E.E., Hanley, J., Bayouth, J., et al. (2009). “Task group 142 doi:10.1016/j.radonc.2022.08.028.
report: Quality assurance of medical accelerators,” Med. Phys. Lewis, B.C., Guta, A., Mackey, S., et al. (2021a). “Evaluation of
36, 4197–4212. doi:10.1118/1.3190392. diffusion-weighted MRI and geometric distortion on a 0.35T
Klüter, S. (2019). “Technical design and concept of a 0.35 MR-LINAC at multiple gantry angles,” J. Appl. Clin. Med.
T MR-linac,” Clin. Transl. Radiat. Oncol. 18, 98–101. Phys. 22, 118–125. doi:10.1002/acm2.13135.
doi:10.1016/j.ctro.2019.04.007. Lewis, B.C., Shin, J., Quinn, B., et al. (2021b). “First clinical expe-
Klüter, S., Schrenk, O., Renkamp, C.K., et al. (2021). “A practical rience of correcting phantom-based image distortion related
implementation of risk management for the clinical introduc- to gantry position on a 0.35T MR-linac,” J. Appl. Clin. Med.
tion of online adaptive magnetic resonance-guided radiother- Phys. 22, 21–28. doi:10.1002/acm2.13404.
apy,” Phys. Imaging Radiat. Oncol. 17, 53–57. doi:10.1016/j. Lim, K., Stewart, J., Kelly, V., et al. (2014). “Dosimetrically trig-
phro.2020.12.005. gered adaptive intensity modulated radiation therapy for cer-
Köhler, M., Vaara, T., Van Grootel, M., et al. (2015). “MR-only vical cancer,” Int. J. Radiat. Oncol. Biol. Phys. 90, 147–154.
simulation for radiotherapy planning,” White paper: Phillips doi:10.1016/j.ijrobp.2014.05.039.
MRCAT for prostate dose calculations using only MRI data. Lim, S.B., Godoy Scripes, P., Napolitano, M., et al. (2021). “An
https://www.documents.philips.com/doclib/enc/fetch/2000/ investigation of using log-file analysis for automated patient-
4504/577242/577251/587787/White_Paper_MR-only_sim_ specific quality assurance in MRgRT,” J. Appl. Clin. Med.
LR.pdf. Phys. 22, 183–188. doi:10.1002/acm2.13361.
Kontaxis, C., Bol, G.H., Stemkens, B., et al. (2017). “Towards fast Lim, S.N., Ahunbay, E.E., Nasief, H., et al. (2020). “Indications of
online intrafraction replanning for free-breathing stereotactic online adaptive replanning based on organ deformation,” Pract.
body radiation therapy with the MR-linac,” Phys. Med. Biol. Radiat. Oncol. 10, e95–e102. doi:10.1016/j.prro.2019.08.007.
62, 7233–7248. doi:10.1088/1361-6560/aa82ae. Ling, C.C., Humm, J., Larson, S., et al. (2000). “Towards multi-
Kontaxis, C., de Muinck Keizer, D.M., Kerkmeijer, L.G.W., et al. dimensional radiotherapy (MD-CRT): Biological imaging and
(2020). “Delivered dose quantification in prostate radio- biological conformality,” Int. J. Radiat. Oncol. Biol. Phys. 47,
therapy using online 3D cine imaging and treatment log files 551–560. doi:10.1016/s0360-3016(00)00467-3.
on a combined 1.5T magnetic resonance imaging and linear Liu, P.Z.Y., Dong, B., Nguyen, D.T., et al. (2020). “First experi-
accelerator system,” Phys. Imaging Radiat. Oncol. 15, 23–29. mental investigation of simultaneously tracking two indepen-
doi:10.1016/j.phro.2020.06.005. dently moving targets on an MRI-linac using real-time MRI
Kontaxis, C., Woodhead, P.L., Bol, G.H., Lagendijk, J.J.W., and and MLC tracking,” Med. Phys. 47, 6440–6449. doi:10.1002/
Raaymakers, B.W. (2021). “Proof-of-concept delivery of mp.14536.
intensity modulated arc therapy on the Elekta Unity 1.5 T Lustig, M., Donoho, D., and Pauly, J.M. (2007). “Sparse MRI:
MR-linac,” Phys. Med. Biol. 66, 04LT01. doi:10.1088/1361- The application of compressed sensing for rapid MR imaging,”
6560/abd66d. Magn. Reson. Med. 58, 1182–1195. doi:10.1002/mrm.21391.
Kooreman, E.S., van Houdt, P.J., Keesman, R., et al. (2020). “ADC Luterstein, E., Cao, M., Lamb, J.M., et al. (2020). “Clinical out-
measurements on the Unity MR-linac: A recommendation on comes using magnetic resonance-guided stereotactic body
behalf of the Elekta Unity MR-linac consortium,” Radiother. radiation therapy in patients with locally advanced cholangio-
Oncol. 153, 106–113. doi:10.1016/j.radonc.2020.09.046. carcinoma,” Adv. Radiat. Oncol. 5, 189–195. doi:10.1016/j.
Kooreman, E.S., van Houdt, P.J., Nowee, M.E., et al. (2019). adro.2019.09.008.
“Feasibility and accuracy of quantitative imaging on a 1.5 Ma, T.M., Ballas, L.K., Wilhalme, H., et al. (2022). “Quality-of-
T MR-linear accelerator,” Radiother. Oncol. 133, 156–162. life outcomes and toxicity profile among patients with local-
doi:10.1016/j.radonc.2019.01.011. ized prostate cancer after radical prostatectomy treated with
Kupelian, P.A., Lee, C., Langen, K.M., et al. (2008). “Evaluation stereotactic body radiation: The SCIMITAR multi-center
of image-guidance strategies in the treatment of localized pros- phase 2 trial,” Int. J. Radiat. Oncol. Biol. Phys. in press (11pp).
tate cancer,” Int. J. Radiat. Oncol. Biol. Phys. 70, 1151–1157. doi:10.1016/j.ijrobp.2022.08.041.
doi:10.1016/j.ijrobp.2007.07.2371. Ma, T.M., Lamb, J.M., Casado, M., et al. (2021a). “Magnetic
Lamb, J., Cao, M., Kishan, A., et al. (2017). “Online adaptive radia- resonance imaging-guided stereotactic body radiotherapy for
tion therapy: Implementation of a new process of care,” Cureus prostate cancer (mirage): A phase III randomized trial,” BMC
9, e1618. doi:10.7759/cureus.1618. Cancer 21, 538. doi:10.1186/s12885-021-08281-x.
96 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Ma, T.M., Neylon, J., Casado, M., et al. (2021b). “Dosimetric Mittauer, K.E., Hill, P.M., Bassetti, M.F., and Bayouth, J.E. (2020).
impact of interfraction prostate and seminal vesicle volume “Validation of an MR-guided online adaptive radiotherapy
changes and rotation: A post-hoc analysis of a phase III ran- (MRgoART) program: Deformation accuracy in a heteroge-
domized trial of MRI-guided versus CT-guided stereotac- neous, deformable, anthropomorphic phantom,” Radiother.
tic body radiotherapy,” Radiother. Oncol. 167, 203–210. Oncol. 146, 97–109. doi:10.1016/j.radonc.2020.02.012.
doi:10.1016/j.radonc.2021.12.037. Mönnich, D., Winter, J., Nachbar, M., et al. (2020). “Quality assur-
Malkov, V.N., and Rogers, D.W.O. (2018). “Monte Carlo study ance of IMRT treatment plans for a 1.5 T MR-linac using a
of ionization chamber magnetic field correction factors as a 2D ionization chamber array and a static solid phantom,” Phys.
function of angle and beam quality,” Med. Phys. 45, 908–925. Med. Biol. 65, 16NT01. doi:10.1088/1361-6560/aba5ec.
doi:10.1002/mp.12716. Murray Brunt, A., Haviland, J.S., Wheatley, D.A., et al. (2020).
Malkov, V.N., and Rogers, D.W.O. (2019). “Erratum: Monte Carlo “Hypofractionated breast radiotherapy for 1 week versus 3
study of ionization chamber magnetic field correction factors weeks (FAST-Forward): 5-year efficacy and late normal tissue
as a function of angle and beam quality [Med. Phys. 45(2), effects results from a multicentre, non-inferiority, randomised,
908–925 (2018)],” Med. Phys. 46, 5367–5370. doi:10.1002/ phase 3 trial,” Lancet 395, 1613–1626. doi:10.1016/s0140-
mp.13782. 6736(20)30932-6.
Malkov, V.N., Hackett, S.L., Wolthaus, J.W.H., Raaymakers, B.W., Mutic, S., and Dempsey, J.F. (2014). “The ViewRay system:
and van Asselen, B. (2019). “Monte Carlo simulations of out- Magnetic resonance-guided and controlled radiotherapy,”
of-field surface doses due to the electron streaming effect in Semin. Radiat. Oncol. 24, 196–199. doi:10.1016/j.semra-
orthogonal magnetic fields,” Phys. Med. Biol. 64, 115029. donc.2014.02.008.
doi:10.1088/1361-6560/ab0aa0. Mutic, S., Palta, J.R., Butker, E.K., et al. (2003). “Quality assur-
Mayinger, M., Kovacs, B., Tanadini-Lang, S., et al. (2020). “First ance for computed-tomography simulators and the com-
magnetic resonance imaging-guided cardiac radioablation of puted-tomography-simulation process: Report of the AAPM
sustained ventricular tachycardia,” Radiother. Oncol. 152, Radiation Therapy Committee Task Group No. 66,” Med.
203–207. doi:10.1016/j.radonc.2020.01.008. Phys. 30, 2762–2792. doi:10.1118/1.1609271.
Mazzola, R., Figlia, V., Rigo, M., et al. (2020). “Feasibility and Nejad-Davarani, S.P., Zakariaei, N., Chen, Y., et al. (2020). “Rapid
safety of 1.5 T MR-guided and daily adapted abdominal-pel- multicontrast brain imaging on a 0.35T MR-linac,” Med. Phys.
vic SBRT for elderly cancer patients: Geriatric assessment 47, 4064–4076. doi:10.1002/mp.14251.
tools and preliminary patient-reported outcomes,” J. Cancer Neph, R., Lyu, Q., Huang, Y., Yang, Y.M., and Sheng, K. (2021).
Res. Clin. Oncol. 146, 2379–2397. doi:10.1007/s00432-020- “DeepMC: A deep learning method for efficient Monte Carlo
03230-w. beamlet dose calculation by predictive denoising in magnetic
McGee, K.P., Hwang, K.P., Sullivan, D.C., et al. (2021). “Magnetic resonance-guided radiotherapy,” Phys. Med. Biol. 66, 035022.
resonance biomarkers in radiation oncology: The report doi:10.1088/1361-6560/abca01.
of AAPM Task Group 294,” Med. Phys. 48, e697–e732. Noel, C.E., Parikh, P.J., Spencer, C.R., et al. (2015). “Comparison
doi:10.1002/mp.14884. of onboard low-field magnetic resonance imaging versus
McNair, H.A., Joyce, E., O’Gara, G., et al. (2021). “Radiographer- onboard computed tomography for anatomy visualization in
led online image guided adaptive radiotherapy: A qualita- radiotherapy,” Acta. Oncol. 54, 1474–1482. doi:10.3109/028
tive investigation of the therapeutic radiographer role,” 4186x.2015.1062541.
Radiography 27, 1085–1093. doi:10.1016/j.radi.2021.04.012. Noel, C.E., Santanam, L., Parikh, P.J., and Mutic, S. (2014).
Meattini, I., Marrazzo, L., Saieva, C., et al. (2020). “Accelerated “Process-based quality management for clinical implemen-
partial-breast irradiation compared with whole-breast irradia- tation of adaptive radiotherapy,” Med. Phys. 41, 081717.
tion for early breast cancer: Long-term results of the random- doi:10.1118/1.4890589.
ized phase III APBI-IMRT-Florence trial,” J. Clin. Oncol. 38, Nowee, M.E., van Pelt, V.W.J., Walraven, I., et al. (2021). “The
4175–4183. doi:10.1200/jco.20.00650. impact of image acquisition time on registration, delineation
Medicines and Healthcare Products Regulatory Agency (2015). and image quality for magnetic resonance guided radiotherapy
“Safety Guidelines for Magnetic Resonance Imaging of prostate cancer patients,” Phys. Imaging Radiat. Oncol. 19,
Equipment in Clinical Use.” Canary Wharf, London, UK. 85–89. doi:10.1016/j.phro.2021.07.002.
Meijsing, I., Raaymakers, B.W., Raaijmakers, A.J., et al. (2009). Oborn, B.M., and Burigo, L.N. (2021). “Monte Carlo modelling
“Dosimetry for the MRI accelerator: The impact of a magnetic in magnetic fields,” pp. 193–208 in Monte Carlo Techniques
field on the response of a Farmer NE2571 ionization cham- in Radiation Therapy, Seco, J., and Verhaegen, F., Eds. (CRC
ber,” Phys. Med. Biol. 54, 2993–3002. doi:10.1088/0031- Press, Boca Raton, FL).
9155/54/10/002. Oborn, B.M., Dowdell, S., Metcalfe, P.E., et al. (2017). “Future of
Menten, M.J., Mohajer, J.K., Nilawar, R., et al. (2020). “Automatic medical physics: Real-time MRI-guided proton therapy,” Med.
reconstruction of the delivered dose of the day using MR-linac Phys. 44, e77–e90. doi:10.1002/mp.12371.
treatment log files and online MR imaging,” Radiother. Oncol. Oborn, B.M., Ge, Y., Hardcastle, N., Metcalfe, P.E., and Keall,
145, 88–94. doi:10.1016/j.radonc.2019.12.010. P.J. (2016). “Dose enhancement in radiotherapy of small lung
Mickevicius, N.J., Kim, J.P., Zhao, J., et al. (2021). “Toward tumors using inline magnetic fields: A Monte Carlo based
magnetic resonance fingerprinting for low-field MR-guided planning study,” Med. Phys. 43, 368. doi:10.1118/1.4938580.
radiation therapy,” Med. Phys. 48, 6930–6940. doi:10.1002/ Oborn, B.M., Metcalfe, P.E., Butson, M.J., and Rosenfeld, A.B.
mp.15202. (2009). “High resolution entry and exit Monte Carlo dose
 References 97

calculations from a linear accelerator 6 MV beam under the Park, J.M., Park, S.Y., Kim, H.J., et al. (2016). “A comparative
influence of transverse magnetic fields,” Med. Phys. 36, 3549– planning study for lung SABR between tri-Co-60 magnetic
3559. doi:10.1118/1.3157203. resonance image guided radiation therapy system and volumet-
Oborn, B.M., Metcalfe, P.E., Butson, M.J., and Rosenfeld, A.B. ric modulated arc therapy,” Radiother. Oncol. 120, 279–285.
(2010). “Monte Carlo characterization of skin doses in 6 MV doi:10.1016/j.radonc.2016.06.013.
transverse field MRI-linac systems: Effect of field size, sur- Pathmanathan, A.U., McNair, H.A., Schmidt, M.A., et al. (2019).
face orientation, magnetic field strength, and exit bolus,” Med. “Comparison of prostate delineation on multimodality imaging
Phys. 37, 5208–5217. doi:10.1118/1.3488980. for MR-guided radiotherapy,” Br. J. Radiol. 92, 20180948.
Oborn, B.M., Metcalfe, P.E., Butson, M.J., Rosenfeld, A.B., and Pathmanathan, A.U., van As, N.J., Kerkmeijer, L.G.W., et al.
Keall, P.J. (2012). “Electron contamination modeling and (2018). ‘Magnetic resonance imaging-guided adaptive radia-
skin dose in 6 MV longitudinal field MRIgRT: Impact of tion therapy: A “Game Changer’ for prostate treatment?” Int.
the MRI and MRI fringe field,” Med. Phys. 39, 874–890. J. Radiat. Oncol. Biol. Phys. 100, 361–373. doi:10.1016/j.
doi:10.1118/1.3676181. ijrobp.2017.10.020.
O’Brien, D.J., and Sawakuchi, G.O. (2017). “Monte Carlo study of Paudel, M.R., Kim, A., Sarfehnia, A., et al. (2016). “Experimental
the chamber-phantom air gap effect in a magnetic field,” Med. evaluation of a GPU-based Monte Carlo dose calculation algo-
Phys. 44, 3830–3838. doi:10.1002/mp.12290. rithm in the Monaco treatment planning system,” J. Appl. Clin.
O’Brien, D.J., Dolan, J., Pencea, S., Schupp, N., and Sawakuchi, Med. Phys. 17, 230–241. doi:10.1120/jacmp.v17i6.6455.
G.O. (2018). “Relative dosimetry with an MR-linac: Response Paulson, E.S., Erickson, B., Schultz, C., and Allen Li, X. (2015).
of ion chambers, diamond, and diode detectors for off-axis, “Comprehensive MRI simulation methodology using a
depth dose, and output factor measurements,” Med. Phys. 45, dedicated MRI scanner in radiation oncology for external
884–897. doi:10.1002/mp.12699. beam radiation treatment planning,” Med. Phys. 42, 28–39.
O’Brien, D.J., Roberts, D.A., Ibbott, G.S., and Sawakuchi, G.O. doi:10.1118/1.4896096.
(2016). “Reference dosimetry in magnetic fields: Formalism Pham, J., Cao, M., Yoon, S.M., et al. (2022). “Dosimetric effects of
and ionization chamber correction factors,” Med. Phys. 43, air cavities for MRI-guided online adaptive radiation therapy
4915. doi:10.1118/1.4959785. (MRgART) of prostate bed after radical prostatectomy,” J.
O’Connor, J.P., Aboagye, E.O., Adams, J.E., et al. (2017). “Imaging Clin. Med. 11, 364. doi:10.3390/jcm11020364.
biomarker roadmap for cancer studies,” Nat. Rev. Clin. Oncol. Prior, P., Chen, X., Botros, M., et al. (2016). “MRI-based
14, 169–186. doi:10.1038/nrclinonc.2016.162. IMRT planning for MR-linac: Comparison between CT-
Olaciregui-Ruiz, I., Vivas-Maiques, B., van der Velden, S., and MRI-based plans for pancreatic and prostate cancers,”
et al. (2021). “Automatic dosimetric verification of online Phys. Med. Biol. 61, 3819–3842. doi:10.1088/0031-
adapted plans on the Unity MR-Linac using 3D EPID 9155/61/10/3819.
dosimetry,” Radiother. Oncol. 157, 241–246. doi:10.1016/j. Raaijmakers, A.J., Raaymakers, B.W., and Lagendijk, J.J. (2008).
radonc.2021.01.037. “Magnetic-field-induced dose effects in MR-guided radiother-
Olberg, S., Green, O., Cai, B., et al. (2018). “Optimization of treat- apy systems: Dependence on the magnetic field strength,” Phys.
ment planning workflow and tumor coverage during daily Med. Biol. 53, 909–923. doi:10.1088/0031-9155/53/4/006.
adaptive magnetic resonance image guided radiation therapy Raaymakers, B.W., Jürgenliemk-Schulz, I., Bol, G., et al. (2017).
(MR-IGRT) of pancreatic cancer,” Radiat. Oncol. 13, 51. “First patients treated with a 1.5 T MRI-Linac: Clinical
doi:10.1186/s13014-018-1000-7. proof of concept of a high-precision, high-field MRI guided
Olch, A.J., Gerig, L., Li, H., Mihaylov, I., and Morgan, A. (2014). radiotherapy treatment,” Phys. Med. Biol. 62, L41–L50.
“Dosimetric effects caused by couch tops and immobilization doi:10.1088/1361-6560/aa9517.
devices: Report of AAPM Task Group 176,” Med. Phys. 41, Raaymakers, B.W., Lagendijk, J.J., Overweg, J., et al. (2009).
061501. doi:10.1118/1.4876299. “Integrating a 1.5 T MRI scanner with a 6 MV accelera-
Oleĭnik, I., and Klepper, L. (1973). “Determination of the optimal tor: Proof of concept,” Phys. Med. Biol. 54, N229–N237.
dynamic conditions for irradiating malignant neoplasms and doi:10.1088/0031-9155/54/12/n01.
adaptive radiation treatment,” Med. Radiol. 18, 49–54. Raaymakers, B.W., Raaijmakers, A.J., and Lagendijk, J.J. (2008).
Overweg, J., Raaymakers, B.W., Lagendijk, J.J., and Brown, “Feasibility of MRI guided proton therapy: Magnetic field dose
K.J. (April, 2009). “System for MRI guided radiotherapy,” effects,” Phys. Med. Biol. 53, 5615–5622. doi:10.1088/0031-
International Society for Magnetic Resonance in Medicine 9155/53/20/003.
17th Annual Meeting, Honolulu, HI. RCR (2021). Postoperative Radiotherapy for Breast Cancer:
Palma, D.A., Olson, R., Harrow, S., et al. (2019). “Stereotactic abla- Hypofractionation RCR Consensus Statements (The Royal
tive radiotherapy versus standard of care palliative treatment in College of Radiologists, London, UK).
patients with oligometastatic cancers (SABR-COMET): A ran- Regnery, S., Buchele, C., Piskorski, L., et al. (2022). “SMART
domised, phase 2, open-label trial,” Lancet 393, 2051–2058. ablation of lymphatic oligometastases in the pelvis and abdo-
doi:10.1016/s0140-6736(18)32487-5. men: Clinical and dosimetry outcomes,” Radiother. Oncol.
Parikh, N.R., Lee, P.P., Raman, S.S., et al. (2020). “Time-driven 168, 106–112. doi:10.1016/j.radonc.2022.01.038.
activity-based costing comparison of CT-guided versus Reynolds, M., Rathee, S., and Fallone, B.G. (2017). “Technical
MR-guided SBRT,” JCO Oncol. Pract. 16, e1378–e1385. Note: Ion chamber angular dependence in a magnetic field,”
doi:10.1200/jop.19.00605. Med. Phys. 44, 4322–4328. doi:10.1002/mp.12405.
98 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

Reynoso, F.J., Curcuru, A., Green, O., et al. (2016). “Technical Sgouros, G., Bolch, W.E., Chiti, A., et al. (2021). “ICRU Report 96,
note: Magnetic field effects on Gafchromic-film response in dosimetry-guided radiopharmaceutical therapy,” J. ICRU 21,
MR-IGRT,” Med. Phys. 43, 6552. doi:10.1118/1.4967486. 1–212. doi:10.1177/14736691211060117.
Rippke, C., Schrenk, O., Renkamp, C.K., et al. (2022). “Quality Shi, L., Chen, Q., Barley, S., et al. (2021). “Benchmarking
assurance for on-table adaptive magnetic resonance guided of deformable image registration for multiple anatomic
radiation therapy: A software tool to complement second- sites using digital data sets with ground-truth deformation
ary dose calculation and failure modes discovered in clinical vector fields,” Pract. Radiat. Oncol. 11, 404–414. doi:10.1016/j.
routine,” J. Appl. Clin. Med. Phys. 23, e13523. doi:10.1002/ prro.2021.02.012.
acm2.13523. Shortall, J., Vasquez Osorio, E., Cree, A., et al. (2021). “Inter- and
Roberts, D.A., Sandin, C., Vesanen, P.T., et al. (2021). “Machine intra-fractional stability of rectal gas in pelvic cancer patients
QA for the Elekta Unity system: A report from the Elekta during MRIgRT,” Med. Phys. 48, 414–426. doi:10.1002/
MR-linac consortium,” Med. Phys. 48, e67–e85. doi:10.1002/ mp.14586.
mp.14764. Sibolt, P., Andersson, L.M., Calmels, L., et al. (2021). “Clinical
Robinson, C.G., Samson, P.P., Moore, K.M.S., et al. (2019). “Phase implementation of artificial intelligence-driven cone-beam
I/II trial of electrophysiology-guided noninvasive cardiac computed tomography-guided online adaptive radiotherapy
radioablation for ventricular tachycardia,” Circulation 139, in the pelvic region,” Phys. Imaging Radiat. Oncol. 17, 1–7.
313–321. doi:10.1161/circulationaha.118.038261. doi:10.1016/j.phro.2020.12.004.
Rogowski, P., von Bestenbostel, R., Walter, F., et al. (2021). Singhrao, K., Fu, J., Wu, H.H., et al. (2020). “A novel anthropomor-
“Feasibility and early clinical experience of online adaptive phic multimodality phantom for MRI-based radiotherapy qual-
MR-guided radiotherapy of liver tumors,” Cancers 13, 1523. ity assurance testing,” Med. Phys. 47, 1443–1451. doi:10.1002/
doi:10.3390/cancers13071523. mp.14027.
Rosenberg, S.A., Henke, L.E., Shaverdian, N., et al. (2019). “A Smit, K., Sjöholm, J., Kok, J.G., Lagendijk, J.J., and Raaymakers,
multi-institutional experience of MR-guided liver stereotac- B.W. (2014). “Relative dosimetry in a 1.5 T magnetic field:
tic body radiation therapy,” Adv. Radiat. Oncol. 4, 142–149. An MR-linac compatible prototype scanning water phan-
doi:10.1016/j.adro.2018.08.005. tom,” Phys. Med. Biol. 59, 4099–4109. doi:10.1088/0031-
Rubinstein, A.E., Liao, Z., Melancon, A.D., et al. (2015). “Technical 9155/59/15/4099.
note: A Monte Carlo study of magnetic-field-induced radia- Snyder, J.E., St-Aubin, J., Yaddanapudi, S., et al. (2022). “Reducing
tion dose effects in mice,” Med. Phys. 42, 5510–5516. MRI-guided radiotherapy planning and delivery times via efficient
doi:10.1118/1.4928600. leaf sequencing and segment shape optimization algorithms,”
Rudra, S., Jiang, N., Rosenberg, S.A., et al. (2019). “Using adap- Phys. Med. Biol. 67, 055005. doi:10.1088/1361-6560/ac5299.
tive magnetic resonance image-guided radiation therapy for Speight, R., Dubec, M., Eccles, C.L., et al. (2021a). “IPEM topical
treatment of inoperable pancreatic cancer,” Cancer Med. 8, report: Guidance on the use of MRI for external beam radio-
2123–2132. doi:10.1002/cam4.2100. therapy treatment planning,” Phys. Med. Biol. 66, 055025.
Sahin, B., Zoto Mustafayev, T., Gungor, G., et al. (2019). “First doi:10.1088/1361-6560/abdc30.
500 fractions delivered with a magnetic resonance-guided Speight, R., Tyyger, M., Schmidt, M.A., et al. (2021b). “IPEM
radiotherapy system: Initial experience,” Cureus 11, e6457. Topical Report: An international IPEM survey of MRI use for
doi:10.7759/cureus.6457. external beam radiotherapy treatment planning,” Phys. Med.
Savenije, M.H.F., Maspero, M., Sikkes, G.G., et al. (2020). “Clinical Biol. 66, 075007. doi:10.1088/1361-6560/abe9f7.
implementation of MRI-based organs-at-risk auto-segmenta- Spindeldreier, C.K., Schrenk, O., Bakenecker, A., et al. (2017).
tion with convolutional networks for prostate radiotherapy,” “Radiation dosimetry in magnetic fields with Farmer-type ioniza-
Radiat. Oncol. 15, 104. doi:10.1186/s13014-020-01528-0. tion chambers: Determination of magnetic field correction fac-
Schellhammer, S.M., Gantz, S., Lühr, A., et al. (2018). “Technical tors for different magnetic field strengths and field orientations,”
note: Experimental verification of magnetic field-induced beam Phys. Med. Biol. 62, 6708–6728. doi:10.1088/1361-6560/aa7ae4.
deflection and Bragg peak displacement for MR-integrated Steinmann, A., Alvarez, P., Lee, H., et al. (2020). “MRIgRT head
proton therapy,” Med. Phys. 45, 3429–3434. doi:10.1002/ and neck anthropomorphic QA phantom: Design, develop-
mp.12961. ment, reproducibility, and feasibility study,” Med. Phys. 47,
Schwartz, D.L., Garden, A.S., Thomas, J., et al. (2012). “Adaptive 604–613. doi:10.1002/mp.13951.
radiotherapy for head-and-neck cancer: Initial clinical out- Steinmann, A., O’Brien, D., Stafford, R., et al. (2019). “Investigation
comes from a prospective trial,” Int. J. Radiat. Oncol. Biol. of TLD and EBT3 performance under the presence of 1.5T,
Phys. 83, 986–993. doi:10.1016/j.ijrobp.2011.08.017. 0.35T, and 0T magnetic field strengths in MR/CT visible mate-
SCoR (2020) MRI Guided Radiotherapy (Society and College of rials,” Med. Phys. 46, 3217–3226. doi:10.1002/mp.13527.
Radiographers, London, UK) https://www.sor.org/learning- Stemkens, B., Paulson, E.S., and Tijssen, R.H.N. (2018). “Nuts
advice/professional-body-guidance-and-publications/docu- and bolts of 4D-MRI for radiotherapy,” Phys. Med. Biol. 63,
ments-and-publications/policy-guidance-document-library/ 21TR01. doi:10.1088/1361-6560/aae56d.
mri-guided-radiotherapy Terpstra, M.L., Maspero, M., Bruijnen, T., et al. (2021). “Real-
Schweikard, A., Shiomi, H., and Adler, J. (2004). “Respiration time 3D motion estimation from undersampled MRI using
tracking in radiosurgery,” Med. Phys. 31, 2738–2741. multi-resolution neural networks,” Med. Phys. 48, 6597–6613.
doi:10.1118/1.1774132. doi:10.1002/mp.15217.
 References 99

Tetar, S.U, Bruynzeel, A.M.E., Bakker, R., et al. (2018). “Patient- van Houdt, P.J., Yang, Y., and van der Heide, U.A. (2021b).
reported outcome measurements on the tolerance of magnetic “Quantitative magnetic resonance imaging for biologi-
resonance imaging-guided radiation therapy,” Cureus 10, cal image-guided adaptive radiotherapy,” Front. Oncol. 10,
e2236. doi:10.7759/cureus.2236. 615643. doi:10.3389/fonc.2020.615643.
Tetar, S.U., Bruynzeel, A.M.E., Lagerwaard, F.J., et al. (2019). van Sörnsen de Koste, J.R., Palacios, M.A., Bruynzeel, A.M.E.,
“Clinical implementation of magnetic resonance imaging et al. (2018). “MR-guided gated stereotactic radiation therapy
guided adaptive radiotherapy for localized prostate can- delivery for lung, adrenal, and pancreatic tumors: A geomet-
cer,” Phys. Imaging Radiat. Oncol. 9, 69–76. doi:10.1016/j. ric analysis,” Int. J. Radiat. Oncol. Biol. Phys. 102, 858–866.
phro.2019.02.002. doi:10.1016/j.ijrobp.2018.05.048.
Thorwarth, D., Ege, M., Nachbar, M., et al. (2020). “Quantitative van Timmeren, J.E., Chamberlain, M., Krayenbuehl, J., et al.
magnetic resonance imaging on hybrid magnetic resonance lin- (2021). “Comparison of beam segment versus full plan re-
ear accelerators: Perspective on technical and clinical valida- optimization in daily magnetic resonance imaging-guided
tion,” Phys. Imaging Radiat. Oncol. 16, 69–73. doi:10.1016/j. online-adaptive radiotherapy,” Phys. Imaging Radiat. Oncol.
phro.2020.09.007. 17, 43–46. doi:10.1016/j.phro.2021.01.001.
Tijssen, R.H.N., Philippens, M.E.P., Paulson, E.S., et al. (2019). Vasmel, J.E., Charaghvandi, R.K., Houweling, A.C., et al. (2020).
“MRI commissioning of 1.5T MR-linac systems: A multi-insti- “Tumor response after neoadjuvant magnetic resonance
tutional study,” Radiother. Oncol. 132, 114–120. doi:10.1016/j. guided single ablative dose partial breast irradiation,” Int.
radonc.2018.12.011. J. Radiat. Oncol. Biol. Phys. 106, 821–829. doi:10.1016/j.
Tocco, B.R., Kishan, A.U., Ma, T.M., Kerkmeijer, L.G.W., and ijrobp.2019.11.406.
Tree, A.C. (2020). “MR-guided radiotherapy for prostate can- Veiga, C., McClelland, J., Moinuddin, S., et al. (2014). “Toward adap-
cer,” Front. Oncol. 10, 616291. doi:10.3389/fonc.2020.616291. tive radiotherapy for head and neck patients: Feasibility study on
Tree, A.C., Khoo, V.S., Eeles, R.A., et al. (2013). “Stereotactic using CT-to-CBCT deformable registration for ‘dose of the day’
body radiotherapy for oligometastases,” Lancet Oncol. 14, calculations,” Med. Phys. 41, 031703. doi:10.1118/1.4864240.
e28–e37. doi:10.1016/s1470-2045(12)70510-7. Verkooijen, H.M., Kerkmeijer, L.G.W., Fuller, C.D., et al. (2017).
Tyagi, N., Fontenla, S., Zhang, J., et al. (2017). “Dosimetric and “R-IDEAL: A framework for systematic clinical evaluation of
workflow evaluation of first commercial synthetic CT software technical innovations in radiation oncology,” Front. Oncol. 7,
for clinical use in pelvis,” Phys. Med. Biol. 62, 2961–2975. 59. doi:10.3389/fonc.2017.00059.
doi:10.1088/1361-6560/aa5452. Vicini, F.A., Cecchini, R.S., White, J.R., et al. (2019). “Long-term
van Dams, R., Wu, T.C., Kishan, A.U., et al. (2022). “Ablative primary results of accelerated partial breast irradiation after
radiotherapy for liver tumors using stereotactic MRI-guidance: breast-conserving surgery for early-stage breast cancer: A ran-
A prospective phase I trial,” Radiother. Oncol. 170, 14–20. domised, phase 3, equivalence trial,” Lancet 394, 2155–2164.
doi:10.1016/j.radonc.2021.06.005. doi:10.1016/s0140-6736(19)32514-0.
van de Lindt, T.N., Fast, M.F., van Kranen, S.R., et al. (2019). Videtic, G.M., Hu, C., Singh, A.K., et al. (2015). “A randomized
“MRI-guided mid-position liver radiotherapy: Validation of phase 2 study comparing 2 stereotactic body radiation therapy
image processing and registration steps,” Radiother. Oncol. schedules for medically inoperable patients with stage I periph-
138, 132–140. doi:10.1016/j.radonc.2019.06.007. eral non-small cell lung cancer: NRG Oncology RTOG 0915
van de Lindt, T.N, Sonke, J.J., Nowee, M., et al. (2018). “A self- (NCCTG N0927),” Int. J. Radiat. Oncol. Biol. Phys. 93, 757–
sorting coronal 4D-MRI method for daily image guidance of 764. doi:10.1016/j.ijrobp.2015.07.2260.
liver lesions on an MR-linac,” Int. J. Radiat. Oncol. Biol. Phys. Voroney, J.P., Brock, K.K., Eccles, C., Haider, M., and Dawson,
102, 875–884. doi:10.1016/j.ijrobp.2018.05.029. L.A. (2006). “Prospective comparison of computed tomogra-
van den Wollenberg, W., de Ruiter, P., Nowee, M.E., et al. (2019). phy and magnetic resonance imaging for liver cancer delinea-
“Investigating the impact of patient arm position in an MR-linac tion using deformable image registration,” Int. J. Radiat. Oncol.
on liver SBRT treatment plans,” Med. Phys. 46, 5144–5151. Biol. Phys. 66, 780–791. doi:10.1016/j.ijrobp.2006.05.035.
doi:10.1002/mp.13826. Wang, J., Salzillo, T., Jiang, Y., et al. (2021). “Stability of
van der Ree, M.H., Blanck, O., Limpens, J., et al. (2020). “Cardiac MRI contrast agents in high-energy radiation of a 1.5T
radioablation—A systematic review,” Heart Rhythm 17, 1381– MR-Linac,” Radiother. Oncol. 161, 55–64. doi:10.1016/j.
1392. doi:10.1016/j.hrthm.2020.03.013. radonc.2021.05.023.
van Heijst, T.C., den Hartogh, M.D., Lagendijk, J.J., van den Wang, Y., Mazur, T.R., Green, O., et al. (2016). “A GPU-
Bongard, H.J., and van Asselen, B. (2013). “MR-guided breast accelerated Monte Carlo dose calculation platform and its
radiotherapy: Feasibility and magnetic-field impact on skin application toward validating an MRI-guided radiation therapy
dose,” Phys. Med. Biol. 58, 5917–5930. doi:10.1088/0031- beam model,” Med. Phys. 43, 4040. doi:10.1118/1.4953198.
9155/58/17/5917. Weiss, E., and Hess, C.F. (2003). “The impact of gross tumor vol-
van Houdt, P.J., Saeed, H., Thorwarth, D., et al. (2021a). ume (GTV) and clinical target volume (CTV) definition on
“Integration of quantitative imaging biomarkers in clinical the total accuracy in radiotherapy,” Strahlenther. Onkol. 179,
trials for MR-guided radiotherapy: Conceptual guidance for 21–30. doi:10.1007/s00066-003-0976-5.
multicentre studies from the MR-Linac Consortium Imaging Werensteijn-Honingh, A.M., Kroon, P.S., Winkel, D., et al.
Biomarker Working Group,” Eur. J. Cancer. 153, 64–71. (2019). “Feasibility of stereotactic radiotherapy using a 1.5 T
doi:10.1016/j.ejca.2021.04.041. MR-linac: Multi-fraction treatment of pelvic lymph node
100 ICRU Report 97, MRI-guided Radiation Therapy    Journal of the ICRU 22(1)1–100

oligometastases,” Radiother. Oncol. 134, 50–54. doi:10.1016/j. radiation therapy (MRgRT) Linac,” J. Appl. Clin. Med. Phys.
radonc.2019.01.024. 21, 127–135. doi:10.1002/acm2.12799.
Westerhoff, J., Daamen, L., Christodouleas, J., et al. (2022). Yan, D., Wong, J., Vicini, F., et al. (1997). “Adaptive modifica-
“Patterns of care and safety in 1800 patients treated on a tion of treatment planning to minimize the deleterious effects
high-field MR-Linac platform registry,” Radiother. Oncol. of treatment setup errors,” Int. J. Radiat. Oncol. Biol. Phys. 38,
170, S363–S364. doi:10.1016/S0167-8140(22)02555-5. 197–206. doi:10.1016/s0360-3016(97)00229-0.
Westley, R., Hall, E., and Tree, A. (2022). “HERMES: Delivery of Yang, Y., Cao, M., Sheng, K., et al. (2016). “Longitudinal diffusion
a speedy prostate cancer treatment,” Clin. Oncol. 34, 426–429. MRI for treatment response assessment: Preliminary experi-
doi:10.1016/j.clon.2022.01.003. ence using an MRI-guided tri-cobalt 60 radiotherapy system,”
Willigenburg, T., de Muinck Keizer, D.M., Peters, M., et al. (2021). Med. Phys. 43, 1369–1373. doi:10.1118/1.4942381.
“Evaluation of daily online contour adaptation by radiation Yang, Y.M., Geurts, M., Smilowitz, J.B., Sterpin, E., and Bednarz,
therapists for prostate cancer treatment on an MRI-guided B.P. (2015). “Monte Carlo simulations of patient dose per-
linear accelerator,” Clin. Transl. Radiat. Oncol. 27, 50–56. turbations in rotational-type radiotherapy due to a transverse
doi:10.1016/j.ctro.2021.01.002. magnetic field: A tomotherapy investigation,” Med. Phys. 42,
Winkel, D., Bol, G.H., Kroon, P.S., et al. (2019). “Adaptive radio- 715–725. doi:10.1118/1.4905168.
therapy: The Elekta Unity MR-linac concept,” Clin. Transl. Zeidan, O.A., Langen, K.M., Meeks, S.L., et al. (2007). “Evaluation
Radiat. Oncol. 18, 54–59. doi:10.1016/j.ctro.2019.04.001. of image-guidance protocols in the treatment of head and
Winkel, D., Bol, G.H., Werensteijn-Honingh, A.M., et al. (2020). neck cancers,” Int. J. Radiat. Oncol. Biol. Phys. 67, 670–677.
“Target coverage and dose criteria based evaluation of the doi:10.1016/j.ijrobp.2006.09.040.
first clinical 1.5T MR-linac SBRT treatments of lymph node Zelefsky, M.J., Kollmeier, M., Cox, B., et al. (2012). “Improved
oligometastases compared with conventional CBCT-linac clinical outcomes with high-dose image guided radiotherapy
treatment,” Radiother. Oncol. 146, 118–125. doi:10.1016/j. compared with non-IGRT for the treatment of clinically local-
radonc.2020.02.011. ized prostate cancer,” Int. J. Radiat. Oncol. Biol. Phys. 84,
Witt, J.S., Rosenberg, S.A., and Bassetti, M.F. (2020). “MRI- 125–129. doi:10.1016/j.ijrobp.2011.11.047.
guided adaptive radiotherapy for liver tumours: Visualising Ziegenhein, P., Pirner, S., Ph Kamerling, C., and Oelfke, U.
the future,” Lancet Oncol. 21, e74–e82. doi:10.1016/s1470- (2015). “Fast CPU-based Monte Carlo simulation for radio-
2045(20)30034-6. therapy dose calculation,” Phys. Med. Biol. 60, 6097–6111.
Woodings, S.J., de Vries, J.H.W., Kok, J.M.G., et al. (2021). doi:10.1088/0031-9155/60/15/6097.
“Acceptance procedure for the linear accelerator component Zurl, B., Tiefling, R., Winkler, P., Kindl, P., and Kapp, K.S.
of the 1.5 T MRI-linac,” J. Appl. Clin. Med. Phys. 22, 45–59. (2014). “Hounsfield units variations: Impact on CT-density
doi:10.1002/acm2.13068. based conversion tables and their effects on dose distribution,”
Yadav, P., Hallil, A., Tewatia, D., Dunkerley, D.A.P., and Paliwal, Strahlenther. Onkol. 190, 88–93. doi:10.1007/s00066-013-
B. (2020). “MOSFET dosimeter characterization in MR-guided 0464-5.