This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016

GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017 1

3D Position Tracking using On-chip Magnetic

Sensing in Image-guided Navigation
Bronchoscopy
Manish Srivastava, Member, IEEE, Kilian O’Donoghue, Aleksandr Sidun, H. Alexander Jaeger,
Alessandro Ferro, Member, IEEE, Daragh Crowley, Member, IEEE, Christian van den Bosch, Marcus
Kennedy, Daniel O’Hare, Member, IEEE and Padraig Cantillon-Murphy, Senior Member, IEEE

Abstract— This paper presents a compact and low-cost I. I NTRODUCTION

on-chip sensor and readout circuit. The sensor achieves
high-resolution 5-degrees-of-freedom (DoF) tracking (x, y,
A. Instrument Position in Clinical Interventions
z, yaw, and pitch). With the help of an external wire wound
sensor, it can also achieve high-resolution 6-degrees-of-
freedom (DoF) tracking (x, y, z, yaw, pitch, and roll angles).
T HE need to know the position and orientation of in-
struments during clinical interventions has always been
critical to successful clinical outcomes. In conventional open
The sensor uses low-frequency magnetic fields to detect
the position and orientation of instruments, providing a vi- surgery, this knowledge was captured by the eyes and hands of
able alternative to using X-rays in image-guided surgery. To the surgeon whose experience and knowledge enabled a virtual
measure the local magnetic field, a highly miniaturised on- reconstruction of the patient’s anatomy albeit in the confines
chip magnetic sensor capable of sensing the magnetic field of the surgeon’s imagination. The advent of laparoscopic
has been developed incorporating an on-chip magnetic surgery and minimally-invasive procedures in the 1980s was
sensor coil, analog-front end, continuous-time ∆Σ analog-
to-digital converter (ADC), LVDS transmitter, bandgap ref- enabled through the addition of more ’eyes;’ the laparoscopic
erence, and voltage regulator. The microchip is fabricated camera which was later augmented with robotic assistance to
using 65 nm CMOS technology and occupies an area of provide stereoscopic views [1]. Now, the position and orien-
1.06 mm2 , the smallest reported among similar designs to tation (pose) of instruments could be visually verified with
the best of our knowledge. The 5-DoF system accurately smaller incisions resulting in faster patient recovery times [2].
navigates with a precision of 1.1 mm within the volume-of-
intrest (VOI) of 15×15×15 cm3 . The 6-DoF system achieves
However, the ‘reconstruction’ of the various camera and line-
a navigation accuracy of 0.8 mm and an angular error of 1.1 of-sight inputs to ‘solve’ instrument poses remained a virtual
degrees in the same VOI. These results were obtained at task in the mind of the surgeon. This complex visualisation
a 20 Hz update rate in benchtop characterisation. The pro- task coupled with the technical challenges of laparoscopic
totype sensor demonstrates accurate position tracking in interventions inevitably limited the scope of early laparoscopic
real-life pre-clinical in-vivo settings within the porcine lung
of a live swine, achieving a reported worst-case registration
interventions to the most common interventions [1].
accuracy of 5.8 mm.
Index Terms— bronchoscopy, electromagnetic tracking,
B. The Growth in Image-Guided Procedures
image-guided interventions, magnetic sensor readout, nav- As an increasing number of surgeons became adept at
igational bronchoscopy, on-chip magnetic sensor, robotic laparoscopic procedures throughout the 1980s and 1990s,
bronchoscopy, robotic surgery, 5-degree-of-freedom, 6-
two important developments greatly simplified the problem
degree-of-freedom.
of instrument pose; the development of robotic assistance in
surgery [3], [4] and the rapid growth in interventions guided
by x-ray imaging [5], [6], particularly in catheter-based proce-
Funded by the European Union ERC-2020-COG Award dures [7]–[9] and orthopedics [10], [11]. Now, instrument pose
101002225. Views and opinions expressed are however those of
the authors only and do not necessarily reflect those of the European could be solved explicitly and objectively using the mechanical
Union or the European Research Council. Neither the European Union transformations from the robotic arm or using the intraop-
nor the granting authority can be held responsible for them. erative x-ray which became the solution of choice in many
The authors would like to acknowledge the support of Science
Foundation Ireland (SFI) and the Microelectronic Circuits Centre Ireland minimally-invasive approaches. However, these new technical
(MCCI). solutions were not without a cost. In robotics, the financial
Manish Srivastava, Kilian O’Donoghue, Aleksandr Sidun, H. implication of high tool costs [12] limited initial applications
Alexander Jaeger, Alessandro Ferro, Daragh Crowley, Christian van
den Bosch, Daniel O’Hare and Pádraig Cantillon-Murphy and are with to the higher-margin clinical disciplines of gynecology [13]
Microelectronics Circuits Centre Ireland, Tyndall National Institute and and specialized surgical interventions for diseases such as
University College Cork, Lee Maltings Complex, Dyke Parade, Cork, T12 prostate cancer [14]. The cost of increased radiation exposure
R5CP, Ireland (e-mail:[email protected]).
Marcus Kennedy is with Cork University Hospital, Wilton, Cork, was human as well as financial. Technical developments in
Ireland. x-ray technology such as three-dimensional x-ray (computed

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
2 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

tomography) and contrast-enhanced fluoroscopy resulted in

dramatic increases in use, particularly in interventional car-
diology [15]–[17], where minimally-invasive ‘image-guided’
procedures (i.e., instrument pose determined from the x-ray
image) transformed common interventions such as vascular
stenting and electrophysiology mapping of the heart from open
surgical interventions to catheter-based day procedures. The
troubling consequences of such rapid and unforeseen growth in
ionizing radiation exposure were acknowledged by the FDA in
their 2010 white paper which stated that “each patient should
get the right imaging exam, at the right time, with the right
radiation dose” [5]. However, the need to resolve the pose of
instruments inside the body, particularly once they moved out
of the camera’s line of sight, became ever more prescient [18]
as ‘image-guided interventions’ (IGI) became the new norm
for minimally invasive procedures [19]. Fig. 1. (a) Pre-operative CT image. (b) Image segmentation. (c) Image
registration to the patient. (d) Image-guided navigation bronchoscopy.
(e) Endoscopic targeting of outer airways.
C. The Clinical Need
As the image-guided interventions became the new norm,
the image was no longer limited to the camera view but could technology has been adopted in high-value minimally-invasive
include ultrasound and optical [20]–[22] images, even from interventions that depend on multiple imaging or robotic inputs
outside the visible spectrum [23], [24], which increased the ne- (e.g., electrophysiology (EP) mapping of the beating heart
cessity to understand the pose of the instrument providing that [43], ear, nose and throat (ENT) surgery [44], and robotic-
image. Contextualizing increasingly complex ‘multi-modal’ assisted bronchoscopy [45]–[47], magnetic tracking remains
image-guided procedures was difficult, if not impossible, in the too expensive for the most common image-guided proce-
clinician’s mind, and image-fusion approaches began to appear dures including interventional bronchoscopy, endoscopy, and
[25]–[27]. Importantly, as concerns grew for the increased use laparoscopic surgery where its benefits may be transformative.
of ionising radiation and the high cost of robotic assistance, For example, an important potential benefit of combining
new technology emerged as critical to the future of image- video with ultrasound in laparoscopic surgery with the help
guided interventions; magnetic tracking [19], [28]. Magnetic of magnetic tracking is more efficient and precise needle
tracking is non-ionising and uses low-frequency electromag- placement during tumor ablation [48]. Magnetic tracking is
netic fields similar to those used in small motors and radio critical in such interventions to accurately ‘register’ the two
controllers to resolve the pose of instruments without any imaging inputs to the patient’s reference frame as indicated
line-of-sight to the instrument [28]. Accuracy and precision in Fig. 1(c). In bronchoscopy, the key advantage of magnetic
can be quickly resolved with sub-millimeter accuracy [19], tracking is that clinicians can combine pre-operative CT with
[29] over the entire volume-of-interest (VOI). However, the the bronchoscope’s video to navigate to the outer airways
technology requires a sensor coil, and sometimes multiple where lung cancer is most likely to originate [49]. The
sensors, inside each instrument to be tracked. Electromagnetic potential to diagnose [50] and treat [51] lung cancer, the
sensors can be miniaturised and the smallest commercial world’s most deadly cancer at an earlier stage by minimally-
devices are less than 500 µm in diameter [30], [31]. One invasive approaches is the key clinical benefit driving a rush
complication of existing magnetic tracking sensors is that by robotic-assisted bronchoscopy systems (e.g., da Vinci Ion,
they suffer from distortion effects which limit accuracy in Auris Monarch and Noah Medical) to integrate magnetic
the presence of metal instruments or equipment [19], [32], tracking technology into their platforms. Sadly, the cost of
[33]. This is usually addressed with intermittent x-ray imaging magnetic tracking sensors remains prohibitive for traditional
[34]–[37], although intelligent mitigation of distortion using manual bronchoscopy, where the camera’s size and limited
sensor fusion [38]–[40] and dynamic model correction [41], field of view prevent access to the outer airways.
[42] appears promising and seems a more plausible long-term
solution if the cost per sensor could be addressed. Fig. 1 shows
the electromagnetic tracking (EMT) system. E. An Integrated Circuit Solution
Motivated by the urgent need for a dramatic reduction in the
D. Limitations of Magnetic Sensor Coils cost per magnetic tracking sensor and the urgent clinical need
Existing electromagnetic tracking (EMT) sensors are neither for reliable diagnostic tools in interventional bronchoscopy,
cheap nor robust. A 5 degree-of-freedom (5-DoF) sensor we have designed, built, and tested the first on-chip magnetic
(which can resolve position as well as yaw and pitch angles, tracking sensor which is compatible with existing magnetic
but not the roll angle needed for many catheter navigation navigation platforms [31] [29] in a pre-clinical airway model,
applications) costs over $25 while 6 degrees-of-freedom (6- successfully navigating to airways well beyond the reach of
DoF) sensor, which provides the additional roll angle in- the bronchoscope’s camera. Recent chip-based solutions [52]
formation are over $250 [31]. This means that while the [53] [54] [55] yield only 3D position coordinates without

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 3

VDD External clock

angular information. Pitch, yaw, and roll angles are critical for BGR LDO_AFE LDO_Digital
Oscillator SCL SDA
clinical applications [56], such as robotic bronchoscopy, which VSS
Power management block
uses angular information to visualize lung airways, and in MUX I2C block

electrophysiology (EP) mapping catheters and biopsy devices, Rcoil

CT-ΔΣ LVDS Doutp
which require extremely small sensors, typically less than Lcoil CCIA PGA Modulator Transmitter Doutn
0.5 mm in diameter. Previous integrated chip-based solutions
Communication block
[52] [54] [55] have failed to integrate on-chip sensing coils, On-chip sensor Analog Front End Data acquisition unit

resulting in large footprints offering little advantage in form On-chip magnetic sensor overview
factor and scaling compared to discrete solutions. The high-
frequency system of Rustom et al. [55] was demonstrated in a Fig. 2. Complete on-chip magnetic sensor overview with a block-level
description of the integrated circuits sub-system.
laboratory setting, whereas in clinical scenarios the use of high
frequencies causes eddy currents in nearby metallic devices, 0.45 mm 1.2 mm
including medical equipment, surgical tables, ventilators, etc. Metal layer Number of turns Width (μm) Spacing (μm)
For this reason, all clinically approved systems using EMT M10 15 7 3
M9 38 2 2
(e.g., NDI Aurora, Medtronic superDimension, Biosense Web- M8 28 5 0.4

ster Carto 3D) operate below 50 kHz to limit interference M10

with nearby metallic devices. In addition to the proposed M9

(a)
M8
solutions, Sharma et al. [53] showcased the integration of
RS
a sensor coil within the chip. However, it is worth noting
that their solution is considerably larger than ours, particularly RSi Csensor
when the commercially available sensor requires a diameter of LS Cm
less than 0.5 mm. Furthermore, their work reported a higher (b)
noise floor (300 √nVHz
), necessitating much stronger magnetic
fields for successful operation. These larger magnetic fields, Fig. 3. (a) On-chip sensor coil arrangement (3D view) with stacked po-
sitioning of metal layers M10, M9, and M8, as well as a table detailing the
in turn, demand substantially higher power requirements and drawn shape geometry. (b) On-chip sensor coil model for low-frequency
pose challenges in scaling to accommodate large tracking multi-layer stacked inductor where Rsi is the resistance of the substrate,
volumes. In contrast, our solution in this study has been the resistance of the metal trace Rs , layer underpass capacitance Cs ,
Cm is total capacitance (including layer to layer (C9−10 , C8−9 and
demonstrated effective at distances of up to 25 cm from the C8−10 ) bond-wire (Cbond ), and packaging capacitance (Cpack )) and
field generator. Moreover, the potential for further extension inductance Ls as described in [59].
exists by simply enlarging the size of the field generator coils.
In this work, we have presented a small and cost-effective
sensor that can be seamlessly integrated into existing magnetic data driver, a bandgap reference (BGR) and low-dropout
navigation systems without requiring significant modifications, (LDO) voltage regulator circuits. As shown in Fig. 2, an ex-
while substantially reducing the scalable cost. ternal clock is used as a test feature for ADC characterisation.
In this work, the integrated circuit (IC) integrates a tiny on-
chip air-core inductive coil which is directly connected to a A. On-Chip Sensor Design
low noise and programmable gain analog-front-end (AFE), and
analog-to-digital converter (ADC). The result is the smallest Designing on-chip magnetic sensors necessitates a spe-
and most robust on-chip sensor solution for magnetic tracking cialised optimisation strategy, markedly distinct from conven-
built with immediate application for image-guided airway tional power applications where inductance and power dissi-
interventions. The paper describes the design, simulation, and pation serve as performance metrics [58]. The performance of
testing of the integrated circuit (IC) which was fabricated in the on-chip sensor coil depends upon two critical parameters:
a 65-nm CMOS process, prior to testing with a customised sensitivity (represented P
by ks , in units of volts per Hertz per
N
electromagnetic tracking system, Anser EMT [29]. Laboratory Tesla, where ks = 2π i=1 Ai with Ai signifying the area
accuracy was measured at 0.8±0.5 mm over a 15 × 15 × 15 of a √single turn within the sensor coil) and thermal noise
cm3 volume. The prototype IC was integrated into a sealed (nV/ Hz) [59]. Increasing the number of turns in the coil can
flexible polypropylene probe for pre-clinical validation and enhance sensitivity, but it comes at the expense of resistance.
used for navigation in live swine airways. Successful in-vivo The thermal noise contributed by the resistance of the coil
registration to pre-operative CT is achieved using the IC sensor traces Rs as per Johnson–Nyquist formula [60], for a given
and worst-case registration accuracy of 5.8 mm is reported in bandwidth ∆f can be expressed as:
the breathing lung. Vn 2 = Sn (f )∆f = 4kB T Rs ∆f (1)
II. IC D ESIGN where ∆f is the sampling bandwidth in Hz, Sn (f ) is the noise
Fig. 2 presents an overview of the integrated circuit archi- voltage spectral density, kB is the Boltzmann constant and T
tecture [57], which encompasses an on-chip inductive coil, is the absolute temperature in K.
analog-front-end (AFE), a data acquisition unit (2nd order The technology used for on-chip sensor design was based on
continuous-time Delta-Sigma Modulator (CT∆ΣM)), oscilla- the coil design criteria, such as feature size and performance
tor, I2C, a low voltage differential signaling (LVDS) output parameters. CMOS 65-nm’s feature size (60-nm) allows it

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
4 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

to be substituted for the commercially available sensor to

Vout1-
reduce overall size. CMOS 65-nm process technology also Rcoil Vout+
includes 10 different types of metal stacks (M1 to M10 metal Lcoil DL-CCIA PGA To ADC
layers), which gives considerable flexibility for the design of
Vout-
the coil by allowing it to be tailored for performance metrics Vout1+
like sensitivity and thermal noise. Older technologies, such as On-chip sensor fc Analog Front End fc
CMOS 90-nm or 180-nm, have fewer metal stack possibilities
(a)
and more resistive metal traces than CMOS 65-nm. As a SW1 Cd
DL-CCIA SW2 Cd
C-tune PGA
result, CMOS 65-nm technology is a viable, low-cost option Cfb SW3 2*Cd
SW4 4*Cd
for meeting the desired specifications of an on-chip sensor coil. Cfb SW5 12*Cd
C-tune

The implementation of a multi-layered coil stacking approach Cin VP1 200*Cd VPP1
Vin+ VN1 Vout1+ 200*Cd VNN1
for the on-chip sensor enhances its sensitivity by augmenting Gm1
Vout1-
Gm1
Vout-
Cin
200*Cd
the effective cross-sectional area interacting with the magnetic VP2 VN2
Vout1+ Vout1-
VPP2 VNN2
Vout+
Vin-
field, as discussed in [59]. The large number of turns of the Cin 200*Cd
Cfb
sensor coil comes at the cost of increased coil resistance Cfb C-tune
(Rs ) and area. Higher Rs leads to a higher noise floor as (b) (c)
per equation (1). The choice of metal layers was guided by
Gm1 vb
the resistance characteristics of each layer, with resistance
increasing from M10 to M1 metal layers. Given the induced Rpd
VP1
Cb
voltage, ranging from several hundred nV to a few hundred µV, Vout1-
VN1 Vo1 Vo11
the uppermost metal layers (M10, M9, and M8 are arranged in gm1 gm2
a stacked configuration) were employed, as shown in Fig. 3(a) Cb Vo12
VP2 Vout1+
along with its corresponding equivalent lumped circuit model Vo2
VN2
(Fig. 3(b)) as discussed in detail in [59]. This choice sig- Rpd MP1
Rpd
nificantly improves sensitivity while mitigating noise levels. MP2
vb pseudo-resistor
The optimisation has been performed based on the magnetic
(d)
flux density of 150 nT and operating frequency of 20 kHz to
gm1 gm2
match the EMT system requirements [29]. The Optimised coil
configuration is shown in Fig. 3(a). Consequently, the designed Vb1 Vpb M6 Vb1 Vcmfb_1
M7 M8
on-chip sensor coil has a resistance
√ value of 1.3 kΩ, translating
Rpd Rpd Vpb_1
to a noise floor of 4.6 nV/ Hz. This necessitates the design M9 M10
VP1 M1 M2 VP2
of a low-noise analog-front-end (AFE) circuit to achieve the
R1 R2 R4
required signal-to-noise ratio (SNR) to meet system accuracy R3
Vo1 VCM1 Vo2 Vout1- VCM2 Vout1+
requirements. C1 C2
VN1 C3 C4
M3 M4 VN2
Rpd Rpd Vo11 M11 M12 Vo12
B. Analog-Front-End Circuit Design
Vb2 Vcmfb M5
Vb2
To meet the system accuracy requirement, input-referred
√ VCM2
VCM1
noise (IRN) density for AFE was specified ≤ 2 nV/ Hz VCMref Vcmfb_1
VCMref Vcmfb
which is 2× lower than sensor thermal noise. Previously
published work has demonstrated the suitability of capacitive- (e) (f)

coupled instrumentation amplifier (CCIA) topology for low- Fig. 4. (a) Block diagram of on-chip sensor with analog-front-end.
noise circuits [61]–[63]. The CCIA is implemented in a fully (b) DL-CCIA schematic showing dual loop feedback architecture. (c)
differential feedback architecture. The input signal to the PGA schematic with capacitive gain architecture. (d) Gm1 block is
implemented as a cascade of a current reuse amplifier and OTA. (e)
CCIA is extremely small (in the range of a few hundred nV Current-reuse amplifier (gm1 ) schematic. (f) OTA (gm2 ) schematic.
to a few hundred µV). As a result, a high gain is necessary
for the CCIA. A high-gain amplifier can get saturated due to
any input offset. To prevent this, the inputs were capacitive such as thermal and low-frequency flicker noise, must be
coupled to the CCIA. minimised.
The simplified single-ended analog-front-end sensor inter- To reduce thermal noise in the MOSFET, the current reuse
face, as shown in Fig. 4(a), consists of a dual loop CCIA and (inverter-based) topology depicted in Fig.4(e) was imple-
a Programmable Gain Amplifier √(PGA). To meet the chal- mented. The fundamental advantage of this topology is that it
lenging IRN specifications (2nV/ Hz), a chopper-stabilized doubles the effective input-referred transconductance without
dual loop CCIA (DL-CCIA) architecture is proposed as shown increasing the bias current, allowing the IRN to be reduced
in Fig. 4(b). This DL-CCIA (GM 1 ) is implemented as a further and can be expressed as:
cascade of a current-reuse amplifier (gm1 ) and telescopic s
operational transconductance amplifier (OTA) (gm2 ), as shown 4kT γ
vrms,in = (2)
in Fig. 4(d). To achieve the targeted IRN, all noise sources gmp + gmn

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 5

where vrms,in is input-referred noise density of gm1 block, k was generated from the main clock with the help of a clock
is Boltzmann constant, T is temperature in Kelvin, γ is the divider circuit. It was designed to be programmable. In normal
excess noise contribution coefficient, gmp is transconductance operating conditions, ÷ 8 (1.28 MHz) clock frequency was
of PMOS (M1 and M2) and gmn is transconductance of used for chopping. Due to chopping artifacts, ripples will
NMOS (M3 and M4) (Fig. 4(e)). The input pair devices (M1- be created at the output [61]. These chopper ripples are
M4) in the current-reuse topology as shown in Fig.4(e), were attenuated by placing a DC-blocking impedance (Cb ) as shown
kept very large and were biased in weak inversion to efficiently in Fig. 4(d), [61] after the first stage current-reuse amplifier.
realize a trans-conductance of 20.5 mS and 17.6 mS for PMOS The output signals of the CCIA ranged from a few µV to
(M1-M2) and NMOS (M3-M4) respectively and thus reducing a few mV. This was due to the choice of a 20 V /V gain
IRN, as shown in equation (2). However, CCIAs are usually for the CCIA, which was selected for noise considerations.
configured as inverting amplifiers, and so their noise gain will To process these low-amplitude signals through the ADC,
be higher than their signal gain, thus degrading their energy additional amplification of these signals was required. To
efficiency. The signal and noise gain can be calculated as : do so, a programmable gain amplifier (PGA), as shown in
Cin Fig. 4(c), was implemented. It has a 5-bit tunable gain to
Signal Gain = (3) accommodate the varying range of induced voltage throughout
Cf b
the volume-of-interest (VOI) in EMT. The gain range of the
Cin + Cf b + CgsM 1−M 4 PGA is 10 V /V to 50 V /V . The total IRN of AFE will be
N oise Gain = (4) dominated by CCIA and can be expressed as:
Cf b
2
where, Cin is the input capacitance of CCIA, Cf b is the 2 2
vn,P GA
vn,AF E = vn,CCIA + (7)
feedback capacitance of the CCIA and CgsM 1−M 4 is the gate A2CCIA
parasitic capacitance of input pair devices (M1-M4) of the where, vn,AF E is the IRN of the AFE, vn,CCIA is the IRN
CCIA, shown in Fig. 4(e). The IRN will be amplified by the of the CCIA, vn,P GA is the IRN of the PGA and ACCIA is
noise gain factor (α) and can be calculated as: the gain of the√ CCIA. The noise level of the PGA has been
Cin + Cf b + CgsM 1−M 4 set at 10 nV / Hz, optimising power consumption.
N oise Gain F actor(α) = (5)
Cin
C. 2nd Order Adder-Less CT∆ΣM
The input capacitors of the CCIA Cin (see Fig. 4(b)) were set
To digitise the signal from the AFE, an on-chip ADC has
to 11.6 pF to mitigate the noise gain (α) penalty caused by
been implemented, as illustrated in Fig. 2. In this work, a 2nd
the gate parasitic capacitance (Cgs ) of the input pair [64]. The
order 1-bit continuous-time-delta-sigma modulator (CT∆ΣM)
dual-loop architecture enables independent biasing of NMOS
topology has been used. The CT∆ΣM’s are well-suited for this
and PMOS in the CCIA, facilitating the operation of the CCIA
work due to their inherent anti-aliasing properties, resistive
from a 1.2 V supply. The dual-loop architecture also makes
input compatibility, and more relaxed settling requirements.
it possible to optimize the parasitic capacitance (Cgs ) of each
Another notable advantage of the CT∆ΣM is its ability to
input pair separately, which reduces noise gains further and
reduce the power supply demands of peripheral components
optimizes IRN, which can be expressed as:
s and the input buffer. This is particularly advantageous for
4kT γ 4kT γ compact sensor systems, as focused on in this study. However,
vrms,in = αp2 + αn2 (6) it is important to note that using a chopping technique in
gmp + gmn gmp + gmn
the AFE introduces a challenge. It generates output spikes
where, αp is the noise-gain factor for PMOS (M1-M2) devices at a frequency equal to twice the chopping frequency (fc ).
and αn is the noise-gain factor for NMOS (M3-M4) devices, An optimal option would be to adjust the frequency of
shown in Fig. 4(e). Large biasing pseudo-resistors (Rpd ≥1 the chopping to twice the sampling frequency (fs ), but, this
GΩ) were implemented (see Fig. 4(e)) to reduce thermal approach is quite inefficient in terms of power consumption
current noise [65]. These GΩ-level impedance pseudo-resistors as the sampling frequency is set to 10.24 MHz. To address and
were utilised to generate high-pass corner frequencies to be mitigate the aliasing issues caused by chopping, the CT∆ΣM
within 1–10 Hz while also biasing the input pairs (M1-M4) was implemented with a finite-impulse-response (FIR) digital-
of DL-CCIA. The 65-nm CMOS process produces large gate to-analog converter (DAC) which creates spectral nulls at
leakage currents in thin oxide gate devices due to tunneling multiples of fs /M, where M is the number of filter taps. In
through the SiO2 gate. This gate leakage can be reduced by [66], the technique of selecting fc = fs /(2 × M) was used to
using thick-oxide gate transistors. The gate leakage current address and reduce chopping aliasing problems. In this work,
can create noise and offsets with the GΩ bias pseudo-resistors. the CT∆ΣM was implemented with a 4-tap FIR DAC. The
This work employs low gate leakage thick oxide gate devices fc used for this work is 1.28 MHz. This frequency is used
(M1-M4) as input pairs (Fig. 4(e)). The AC coupling and dual to align the notches in the frequency response of the AFE
loop topology allow these thick oxide devices to be operated with the nulls of the FIR feedback DAC. Fig. 5 displays the
from a 1.2 V supply. simplified single-ended schematic of the second-order 1-bit
To reduce the low-frequency flicker noise from the CCIA CT∆ΣM.
and to achieve the challenging IRN, chopping was employed The second integrator serves the dual purpose of being
at the input of CCIA (see Fig. 4(a)). The chopping frequency both an integrator and an adder circuit which uses less area

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
6 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

Duty-cycled resistor(Rfb)

Cin Ck1 Ck2

Rin
R1
Vin Dout
vo1
vo2
R-DAC

CT-PS
FIR DAC

ELDC
-1 -1 -1
D Z Z Z DAC fs
4 4
F(z) CD1 CD2 CD3 CD4 Fc(z)
Z-0.5
Rin = 30 kΩ R1 = 390 kΩ Cin = 17.5 pF Ck2 = 200 fF
Fig. 5. Simplified single-ended implementation of the second-order 1-bit adder-less CT∆ΣM.

Iref clock output (CLK) can be tuned to 10.24 MHz and has a
variance of ± 8.9 kHz with a duty cycle of 51.4%. The FLL
Vref circuit consumes 67.86 µW from a 1.2 power supply and
M1 M2 M3 occupies 0.019 mm2 silicon area. The periodic jitter of the
C1 output clock of FLL exhibits a standard deviation of 32.59 ps
rms.
MP
E. Peripherial Blocks Implementations
C3 To reduce power consumption and minimize output signal
MN
distortion during data transmission, low voltage differential
C2 R1 CLK
signaling (LVDS) data driving buffers were implemented for
the transmission of the digitised sensor data. As the single bit
Fig. 6. Schematic of the frequency-locked-loop clock circuit. CT∆ΣM output is already in a data-encoded serial format,
the undecimated output is applied to LVDS. To implement the
whole system-on-chip, peripheral blocks such as I2C, bandgap
compared to the previously published work [66], [67]. The
voltage reference (BGR), and voltage regulators were also
integrators in the CT∆ΣM were realized by folded cascode
designed, as shown in Fig. 2. The I2C block serves several
OTAs, while the feedback FIR DAC is realized as an equal-
critical functions, including the programming of the gain of
weighted four-tap resistive DAC. This design achieves a state-
the PGA, selection of chopping and FLL frequencies, and con-
of-the-art area of 0.07 mm2 with signal-to-noise-and-distortion
figuration of the integrator RC time constant in the CT∆ΣM.
ratio (SNDR) and dynamic range (DR) values of 81.5 dB and
Additional sub-blocks, depicted in Fig. 2, play vital roles in
91 dB, respectively. The CT∆ΣM consumes 300 µW from a
maintaining the stability and accuracy of on-chip operations.
1.2 V power supply.
For instance, the BGR circuit ensures the availability of stable
reference voltage (0.6 V) and currents for AFE, the CT∆ΣM,
D. Frequency-Locked Loop Circuit the FLL, and the LVDS. Low-dropout (LDO) regulator circuits
To ensure proper timing and synchronization within the were implemented generating a 1.2 V power supply from
system, a customized on-chip frequency-locked loop (FLL) an external 1.8 V power supply. The LDO regulator circuits
clock circuit has been implemented, providing the clock signal have been strategically implemented to address potential dis-
for sampling the CT∆ΣM, facilitating chopping for the AFE, turbances in the power supply, specifically switching noise.
and enabling communication within the I2C block. Fig. 6 Their presence effectively mitigates transient disruptions and
illustrates the schematic of the on-chip frequency-locked loop noise artifacts, thus minimising any adverse impact on system
clock circuit. The FLL design was inspired by [68] and performance due to power supply fluctuations. Two LDOs
voltage-controlled oscillator (VCO) design used in FLL was were implemented as shown in Fig. 2, with one dedicated
referenced from [69]. For this work, the area and power were to powering the AFE and the other responsible for supplying
major constraints for the FLL circuit. The frequency of the power to the digital blocks, including the CT∆ΣM. Using

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 7

U1 U2 C1 between each coil. These frequency intervals were carefully

8 channel signal generator
LFG selected to minimise interference between the adjacent coils.
Vin RDC To generate the required magnetic fields for this system,
CFG
C2 we employed planar windings on a printed circuit board,
C1 = 330 μF, C2 = 22 μF, RDC = 100 kΩ, R which offers a cost-effective and precise method for cre-
S
RS = 1 Ω, LFG = 3.9 mH ating accurate magnetic fields and their associated models
(a) 22 cm for tracking applications. Square coils were chosen for their
(b)
42 cm 8.5 cm simplicity in comparison to circular and other coil formations,
making the necessary calculations more straightforward. When
considering the coil as composed of multiple straight filaments,
the Biot–Savart law was applied to calculate the magnetic
42 cm

field resulting from each filament. A more comprehensive

8.5 cm
exploration of this method can be found in [29]. Each coil
has a side length of 8.5 cm and comprises a total of 255
turns distributed across six copper layers. To minimise coil
resistance, each PCB copper layer measures 105 µm (3 oz
(c) (d) copper plating) in thickness. Fig. 7(c) illustrates the complete
Fig. 7. (a) The power amplifier circuit used to deliver sinusoidal
array of the eight coils utilised in the field generator while
AC current to each of the eight coils used in the field generator coil Fig. 7(d) displays an individual coil. To optimise current
array. U1 is the LM7171 high-speed operational amplifier and U2 is an output on each channel, the coils were fine-tuned using low
LMH6321 current buffer. Capacitance, C1 blocks any DC current from
the coil, CF G tunes the AC impedance of the coil at the chosen transmit
equivalent series resistance (ESR) capacitors to reduce the
frequencies to improve the output current capability. A precision current overall impedance of the coils. The coils used in the field
sense resistor is used to provide feedback on the measured current (b) generator were characterized by a DC resistance of 29.1 Ω
An Electronics board is used to deliver the drive signals to each field
generator change. (c) Eight coils utilised in the field generator. (d) Single
and an inductance of 3.9 mH. When driven at resonance with
coil. a peak current of 150 mA, this results in a peak voltage of
4.4 V. Importantly, this voltage level comfortably falls within
the linear range of the LMH6321, ensuring minimal signal
separate LDOs offers the advantages of noise isolation and distortion. Fig. 7(b) illustrates the electronic circuit board used
signal integrity. for generating signals for the transmitter coil.

III. T RANSMITTER D ESIGN IV. R ESULTS

In this study, the transmitted magnetic field is generated A. IC Characterization

by the EMT system [29], a configuration comprising eight The on-chip magnetic sensor and associated sub-systems, as
transmitter coils. Each of these coils generates an alternating shown in Fig. 2, was implemented in a 65-nm CMOS process
current (AC) magnetic field operating within the 20-34 kHz and occupies an active area of 1.06 mm2 . The on-chip sensor
frequency spectrum. Each coil is driven by an AC amplifier system (including all blocks as shown in Fig. 2) consumes 6
within the field generator’s control unit. Due to the maximum mW power from a 1.8 V power supply and Fig. 8 illustrates the
displacement of the receiver’s sensor being 35-40 cm, the breakdown of power usage of all blocks. Measurement results
magnetic field strength features a high dynamic range within of core blocks such as the on-chip sensor coil, the AFE, and
the transmitter’s VOI due to the inverse-cube relationship of the CT∆ΣM are shown in Fig. 9 and Fig. 11 respectively.
the magnetic field strength with distance from the field source. Impedance measurements for the on-chip coil sensor were
This results in a corresponding dynamic range in the voltage
induced on the sensor coil, resulting in a corresponding vari- TABLE I
ation in the voltage induced by the sensor coil. Consequently, M EASURED ON - CHIP SENSOR PARAMETERS AND
the induced voltage spans a range from several hundred nano-
PERFORMANCE [59]
Volts (nV) at the limits of operation, to a few hundred micro-
Volts (µV), close to the field generator board. On-chip coil parameters estimation
As shown in Fig. 7(a), U1 corresponds to the LM7171 Coil Calculation Measurement
(Texas Instruments) high-speed operational amplifier, while U2 parameter result result
represents the LMH6321 (Texas Instruments) current buffer. Sensitivity, µV /T /Hz 202.3 197.9
The capacitor’s (C1 ) role is to block any direct current (DC) SNRa, dB 23.6 22.1
originating from the coil, and CF G is utilised to fine-tune Inductance, µH 7.7 8.07
the AC impedance of the coil to resonate at selected trans- Resistance, Ω 1276 1317
mit frequencies, thereby enhancing the output current drive a: At f=20kHz and B=150nT
capacity. To accurately measure current, a precision current
sense resistor is employed. The 8 transmit frequencies for each performed using an HP4285A precision LCR meter. Another
coil were configured between 20-34 kHz in intervals of 2 kHz key metric for the on-chip sensor coil design was SNR, as

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
8 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

3%
15%
CCIA
PGA
LDO_Analog
5%
44% LDO_Digital
2% Total Power 6 mW BGR
5% Clock Source
CTDSM
4%
LVDS (a) (b)
4%
Digital
Fig. 9. (a) Simulated and measured AFE input-referred noise density
18% with and without chopping. (b) Simulated and measured DC gain of AFE.

Fig. 8. Power breakdown for various blocks of on-chip sensor.

it represents both the noise and sensitivity of the inductor

[59]. SNR has been measured with the AFE and an off-chip
second-order butterworth filter to prevent aliasing. The data
acquisition has been performed using a NIDAQ USB-6343
(a) (b)
(National Instruments, Austin, TX, USA). The signal was fed
into the on-chip sensor coil by a 3-D Ferronato BH300HF-3-B Fig. 10. (a) Measured CMRR of AFE. (d) Measured PSRR of AFE
without LDO
Helmholtz coil, which is optimized for generating fields with
frequencies from DC up to 300 kHz and field strengths up to
150 µT. The measured results of SNR, sensitivity, inductance,
and resistance of the on-chip sensor coil are presented in
Table I and can be found in more details in our previous work
[59]. Self-resonance frequency of the on-chip sensor coil has
been measured in our previous work [59] as approximately 8
MHz. As the tones to be detected are in an 18 kHz range, the
receiver on-chip coil wasn’t designed to resonate.
(a) (b)
The IRN of the AFE (with and without chopping), was
measured with an audio analyser (SR1 — Dual-domain audio Fig. 11. (a) Power Spectrum Density of CT∆ΣM with input amplitude of
analyser from Stanford Research Systems) while the AFE’s -2.4 dBFS. (b) SNDR over increasing input signal amplitude at 6.17 kHz.
inputs were grounded and the PGA’s gain√was set to maximum.
At 1 kHz, the AFE’s IRN was 2.07 nV / Hz with a chopping
frequency at 1.28 MHz as shown in Fig. 9(a). Fig. 9 (a) shows signal band (20–34) kHz. CCIA is the block most vulnerable
the simulated and measured IRN with and without chopping. to power supply noise due to the small sensor signal. The
The simulated result is a pnoise simulation and measured result on-chip LDO for AFE has more than -50 dB (simulated)
comparable with the simulated result. The AFE achieved a PSRR so the total PSRR would be more than -90 dB in the
maximum DC gain of 55.6 dB within a bandwidth of 100 Hz – signal band which is sufficient for noise considerations. The
60 kHz, featuring a high-pass (HP) filter pole at 2 Hz as shown AFE consumes 2.6 mA and occupies 0.18 mm2 active area.
in Fig. 9(b). It is worth noting that the simulation result was A comparison with the state-of-the-art for CCIA has been
obtained without chopping. As the open loop amplifier gain presented in Table II.
at the chopping frequency is less than the response at close
TABLE II
to DC the measured result should be close to the simulated C OMPARISON WITH THE STATE - OF - THE - ART
result at the chopping frequency of 500 kHz, as shown in Metrics This work [50] [63] [70]
CMOS technology [nm] 65 180 180 180
Fig. 9(b). The peaking in the gain response is caused by the Architecture CCIA CCIA+CTDSM 3-Opam Inverter-stacking
Active Area [mm2 ] 0.18a 0.73 0.22 0.02
AC coupling at the input as shown in Fig. 4(b) and the DL- Supply Voltage [V] 1.2 1.8 1.55 1
CCIA inter-stage AC coupling as shown in Fig. 4(d). The Current [mA] √ 1.48b 1.2 1.3 0.00025
Input Noise Density [nV / Hz] 2.07 3.7 8.2 55
composite bi-quad pole-zero responses create the peaking. Input Offset [µV ] 1.25 7 N.R. N.R.
The common-mode rejection ratio (CMRR) and power-supply CMRR [dB]
NEF/PEFc
82
3.12/11.68
134
5/44
N.R.
11.4/201
82
1.07/1.14
rejection ratio (PSRR) for AFE (CCIA and PGA) without a
CCIA and PGA area included b
CCIA current only c
PEF = VDD*N EF 2 N.R.: Not reported

LDO were measured and shown in Fig. 10 (a) and Fig. 10

(b), respectively. The CMRR was reported more than 80 dB For the CT∆ΣM characterisation, an audio precision signal
in the signal band (20-34 kHz) and PSRR was −40 dB in the generator (Keysight 33500b) was used to apply a fully dif-

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 9

ferential input sine wave. The single-bit digital output data and consistently. The robot was mounted on a wooden frame to
was read out using a Digilent Digital Discovery portable logic keep it more than 1 m from the volume-of-interest (VOI) and
analyser (Digilent) and further off-chip post-processing was avoid magnetic field interference. Planar electromagnetic coils
carried out using MATLAB. A 6.17-kHz sine wave input in the field generator are designed to produce AC magnetic
signal with an amplitude of −2.4 dBFS (full scale of 1.2 fields within a scalable VOI of 30×30×25 cm3 . The IC was
V) was applied to the input of CT∆ΣM while the sampling mounted on a 3 cm×3 cm PCB for benchtop characterization,
clock frequency was set to 10.24 MHz. The output spectrum as illustrated in Fig. 13(b). The 10.24 MHz LVDS output
of the modulator, acquired by an average of eight 130K point data was transmitted via a flexible PCB. The PCB tracks are
Fast-Fourier transforms (FFTs) using a Hanning window, is terminated with a resistance (R1) of 113 Ω. Subsequently, the
shown in Fig. 11(a). The modulator achieves 81.5 dB SNDR, signals were fed to a comparator, followed by a D flip-flop.
with HD3 and HD5 of -98.98 dB and -98.0 dB, respectively. The flip-flop’s output data and clock output signals from the
Fig. 11(b) shows the SNDR for a 6.17 kHz tone versus the chip were sampled at a rate of 100 MSPS using a Digital
input signal amplitude level in 20 kHz bandwidth. Sweeping Discovery portable logic analyzer (Digilent). A discrete-time
the input signal from -100 dBFS to 0 dBFS demonstrates a Fourier transform (DFT) algorithm was employed to extract
dynamic range of 91 dB. The CT∆ΣM occupies an active area the amplitude and phase of the carrier tones. For each pose
of 0.07 mm2 . estimate, 221 samples were collected, with 21 ms required per
This on-chip sensor system will be used in applications data acquisition and 29 ms for pose determination, resulting in
such as magnetic sensing in position tracking for image-guided a fixed update rate of 20 Hz for the system. The accuracy of
or robotic interventions. To demonstrate the performance for
magnetic tracking, system tracking measurements were carried TABLE III
out using the new sensors and the open-hardware EMT system 5-D O F ACCURACY RESULTS OF THE ON - CHIP SENSOR
[29]. As described in Section III, the EMT system uses eight VOI 15×15×15 cm3 25×25×25 cm3
signal tones for tracking the instrument during image-guided Mean Position Accuracy [mm] 1.1 1.5
or robotic interventions [29]. The on-chip sensor achieves RMS Position Accuracy [mm] 1.7 1.9
STD Position Accuracy [mm] 1.1 1.2
digitisation of the eight signal tones within the frequency range
of 20-34 kHz and transmits the desired output off-chip for
further processing to find the position of the instrument [29]. the sensor was evaluated by recording a grid of measurements
The test setup is shown in Fig. 12(a). and comparing them with the actual positions of the sensor
Fig. 12(b) shows the resulting output spectrum of the system evaluated using an optical tracker as a position gold standard
using the Digital Discovery logic analyser. The spectrum has (NDI Polaris, Northern Digital Inc., Waterloo, Canada). A
high-frequency tones (> 40 kHz) due to the chopping applied grid consisting of 405 test points was recorded, organized in
to the AFE which were filtered out off-chip. The spectrum a 9×9×5 configuration, with intervals of 40 mm in the x-y
was analysed by performing an average of eight 130K-point plane and a separation of z = 50 mm between layers. The
Fast-Fourier transforms (FFT). The eight tones exhibit varying tests were repeated with two different sensor orientations: one
signal strengths due to the dependence of induced voltage on in the z-direction and the other in the y-direction, resulting
the magnetic sensor coil’s distance from the eight transmitter in a total of 810 test points. The position error was defined
coils. as the discrepancy between the measured grid of points and
The CT∆ΣM’s output data and clock output signals from the reference grid (optically measured grid) when both grids
the chip were sampled at a rate of 100 MSPS using the Digital were aligned using Horn’s absolute orientation algorithm [71].
Discovery portable logic analyser (Digilent). A discrete-time The sensor operates at a refresh rate of 20 Hz, but for these
Fourier transform (DFT) algorithm was employed to extract measurements, a 2-second averaging period is employed for
the amplitude and phase of the carrier tones as shown in each point to eliminate mechanical vibrations. The measured
Fig. 12(c). These extracted tones are then used to resolve the results of the accuracy experiments are presented in Fig. 14.
position and orientation (pose) of the magnetic sensor used The mean measurements indicate a position error of 1.5 ± 1.2
for the electromagnetic tracking by applying the Levenberg- mm for a 25×25×25 cm3 volume, reducing to 1.1 ± 1.1 mm
Marquardt non-linear least squares method as described in for a 15×15×15 cm3 volume. Table III presents the system
[29]. accuracy with the on-chip 5-DoF sensor.

B. IC 5-DoF Sensor Characterization C. IC Sensor Integration with 6-DoF Implementation

To evaluate the EMT system with the on-chip sensor, ex- The on-chip sensor presented in this study achieves 5-DoF.
tensive tests were conducted using an optical tracking system To achieve precise 3D (6-DoF) pose estimation, the use of
(accuracy <10µm) as a reference [29]. Fig. 13 shows the two 5-DoF magnetic sensors is required, positioned with a
measurement setup. The on-chip sensor presented in this study certain angular displacement between their sensing axes. The
operates in 1D, enabling the system to achieve 5 degrees-of- degree of accuracy and robustness of the estimate is directly
freedom (5-DoF) localisation using the Levenberg-Marquardt proportional to the angle between these sensors. Augmenting
non-linear least squares method outlined in the prior work [29]. the 5-DoF with an additional sensitive ferromagnetic cored
A 2D linear actuator was used to move the sensors repeatedly sensor produces a highly accurate 6-DoF sensor setup that is

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
10 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

Fig. 12. (a) Test setup for measurement of the IC performance. (b) Magnitude plot showing the performance of the CT∆ΣM with magnetic sensor
with the time-varying 8-tones (20-34 kHz) magnetic field. (c) DFT of the output from the CT∆ΣM for the representative 8 magnetic field tones
(20-34 kHz). This is a typical frequency range for frequency-domain magnetic tracking [29].

not only cost-effective but also compact and competitive in the device, a second cross product taking the two perpendicular
terms of form factor. components already known can be computed using (13).
1) Roll Angle Measurement: Consider the device consists of
two 5-DoF sensors rigidly positioned relative to each other n̂ab = n̂a × n̂b (12)
at some angle so as not to be co-linear with respect to
n̂c = n̂ab × n̂b (13)
one another. The positions of the two sensors are given by
coordinates ra and rb , each with an orientation unit vector The next step is to calculate the rigid origin of the device. In
given by n̂a and n̂b respectively, these are shown in Fig. 15. the simplest case, the geometric mean of the two components
can be taken to find the center point. However, considering
ra = (xa , ya , za ) (8)
the case where one sensor is significantly noisier and less
reliable than the other, a weighting factor, γ, can be used
rb = (xb , yb , zb ) (9)
to increase the reliance on the more robust sensor output to
improve the overall accuracy. The weighted position can be
n̂a = (nxa , nya , nza ) (10) calculated using (14).
n̂b = (nxb , nyb , nzb ) (11) rc = γra + (1 − γ)rb = (xc , yc , zc ) (14)

To find the roll angle of the rigid device, the cross product Experimentally, a value of γ of 0.6 was found to give the
of two sensor unit vectors can be used as follows to find best accuracy results. Combining the rotation matrix and the
the vector perpendicular to both sensors, as given in (12). To resulting position information, the following transformation
complete the full rotation matrix to describe the orientation of matrix can be used to fully describe the 6-DOF pose of the

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 11

450 µm

XY Actuator Optical marker

Sensor Coil
1200 µm
BGR

AFE(CCIA and PGA)Digital

3 cm
LDO
3 cm

CTΔΣM
1164 µm
(b) (c)
Field Generator Z
I2C
Y

Oscillator
LVDS
LDO

+
X
(a) (d)
(a)
Fig. 13. Measurement setup (a) The XY robot moves the magnetic
sensor on a horizontal plane with ground truth accuracy verified using
optical tracker markers. A wooden frame is used to avoid magnetic
field distortions. (b) The test PCB with on-chip magnetic sensors. (c)
Enlargement of the magnetic sensor chip and die. (d) Chip Micrograph.

device.  x
nxab nxa

nc xc
nyc nyab nya yc 
T =
nzc
 (15)
nzab nza zc 
0 0 0 1

Within the sensor assembly, a wire-wound sensor, measur-

ing 0.5 mm in diameter and 8 mm in length, is positioned
along the catheter device’s axis. The on-chip sensor is co- (b)
located on the flexible printed circuit board (flex PCB) and
records the magnetic field perpendicular to the plane of the
PCB as shown in Fig. 16. To evaluate the EMT system with
the on-chip sensor, extensive tests were conducted using the
same setup as the 5-DoF measurements (Fig. 13).
The accuracy of both the position and orientation esti-
mates of the sensor was evaluated by recording a grid of
measurements and comparing them with the actual positions
of the sensor evaluated using a 6-DOF optical tracker as a
gold standard (NDI Polaris, Northern Digital Inc., Waterloo,
Canada). A grid consisting of 405 test points was recorded,
organized in a 9×9×5 configuration, with intervals of 40 mm Y
Z
in the x-y plane and a separation of z = 50 mm between layers. X
The tests were repeated with two different sensor orientations:
one in the z-direction and the other in the y-direction, resulting
in a total of 810 test points. The position error was defined
as the discrepancy between the measured grid of points and
the reference grid (optically measured grid) when both grids (c)
were aligned using Horn’s absolute orientation algorithm [65].
The orientation error was calculated in a similar manner by Fig. 14. 5-DoF Measurement results (a) The magnetic field generated
by the 8 transmitter coils was measured at five grids of 81 points for two
comparing the orientation of each test point with the true sensor orientations. (b) Cumulative Distribution Function of the position
orientation as reported by the optical tracking system. error calculated at test points. (c) The accuracy of resolved sensor
position was calculated from 810 test points.
The sensor operates at a refresh rate of 20 Hz. The measured
results of the accuracy experiments are presented in Fig. 17.
The mean measurements indicate a position error of 1.1 ± 0.8
mm and an angular error of 1.8 degrees for a 25×25×25 cm3 degrees for a 15 × 15 × 15 cm3 volume. Table IV presents the
volume, reducing to 0.8 ± 0.5 mm and an angular error of 1.1 system accuracy with the aforementioned VOI.

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
12 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

(a)
Fig. 15. Diagram showing two 5-DoF sensors showing both their
position coordinates and orientation unit vectors. A full rotation matrix
describing the orientation of the device can be computed using two
cross-product calculations. -
5-

Sensing axes
3 mm OD

Chip sensor (a) Flex PCB

External wire-wound sensor

(b)
(b)

(c)

Fig. 16. (a) Diagram showing the structure of the 6-DoF device. An
off-the-shelf 8 mm long, 0.5 mm diameter navigation sensor is bonded
to the flex PCB along with the on-chip sensor. (b) Shows the assembled
device with the on-chip sensor and a wire wound sensor. (c) Flex PCB
sealed in flexible polypropylene probe for pre-clinical validation and used
for navigation in live swine airways.

TABLE IV
6-D O F ACCURACY RESULTS OF THE COMBINED SENSOR ( O N -C HIP
Field generator
AND EXTERNAL WIRE - WOUND

VOI 15×15×15 cm3 25×25×25 cm3

Mean Position Accuracy [mm] 0.8 1.1
RMS Position Accuracy [mm] 0.9 1.3
STD Position Accuracy [mm] 0.5 0.8
Mean Angular Accuracy [Deg] 1.1 1.8
RMS Angular Accuracy [Deg] 1.3 2.1 (c)
STD Angular Accuracy [Deg] 0.6 1.1
Fig. 17. Accuracy results (a) Position accuracy of the 6-DoF sensor
device. Accuracy in 30 mm thick volume slices was averaged to show
the underlying trend with respect to distance from the origin of the
V. P RE -C LINICAL E VALUATION : N AVIGATION field generator. (b) Cumulative distribution of the errors of each sensor
individually and the combination of the two. Combining the two sen-
B RONCHOSCOPY sors gives the advantages of accuracy and also the roll angle of the
device. (c) The magnetic field generated by the 8 transmitter coils was
The IC was evaluated for technical efficacy and performance measured at 980 test points, for two sensor orientations: one in the z-
in a real-life pre-clinical environment of the porcine lung in direction and the other in the y-direction
a similar approach to existing commercial solutions that rely
on inductive coil-based sensing. To facilitate its application,

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 13

the prototype sensor was integrated into a sealed flexible Slicer 5.

Tracked sensor probe Virtual
polypropylene probe as shown in Fig. 16(c), which was endoscopic
view
utilized for navigation within the airways of breathing swine as
shown in Fig. 16. A female swine (sus scrofa, 42kg) was used
in accordance with the European Directive 2010/63 and French
laws concerning animal protection in laboratories. Procedures
were approved by the local Ethical Committee and authorized
by the French Ministry of Education, Research and Innova- Endoscopic camera view Virtual navigation pathway
tion under Protocol notification #28599-2020121012122760.
(a)
Before experimentation the animal was group-housed and
acclimatized for 7 to 10 days in an enriched environment, re-
specting circadian cycles of light-darkness, and with controlled
humidity and temperature conditions. The animals were pellet-
fed twice a day (piglet diet Lorial/Costal 10-20g/Kg/day) and
fasted 24 hours before the anesthesia with ad libitum access
to water. Procedures were performed under general anesthe-
sia in an experimental hybrid operating room. Each animal
is sedated with an intramuscular injection of azaperone1-
2 mg/Kg (Stresnil® Elanco) combined with zolazepam and
tiletamine at 5-10 mg/kg (Zoletil® Virbac) to be transferred to (b) (c)
the operative room. Anesthesia was induced using intravenous
Fig. 18. Pre-clinical results (a) in-vivo registration to pre-operative
propofol 2-4 mg/kg (Propomitor® , Osalia) and rocuronium 1-2 CT using the on-chip magnetic (b) Position measurements recorded
mg/kg (Esmeron® , MSD), followed by orotracheal intubation. and overlayed on segmented airways model of pig used for in-vivo
General anesthesia was maintained with inhaled isoflurane 2% testing (c) Chest displacement over a 100-second window as recorded
using the on-chip magnetic sensor. This data was recorded in 1-second
(Isoflo® , Zoetis) in an Air/O2 mixture. Controlled ventilation increments and filtered using a 5-point moving average filter.
was achieved using the Drager Primus® system, with oximetry
and heart rate monitoring during the procedure. Analgesia was
provided by intravenous injection of buprecare 0.01 mg/kg
[56] as shown in Fig. 18(a). As depicted in Fig. 18(b), the
(Bupaq® , Virbac).
prototype sensor successfully achieved in-vivo registration to
Following anesthesia, the animal’s main airway was dilated
preoperative CT, demonstrating its ability to accurately track
and an 8.5 French tracheal tube (Portex ET tube, Smiths
the position of the bronchoscope where the blue points repre-
Medical, UK) was inserted with a proximal Y-connector for
sent the airway centerline, while the orange markers indicate
continuous ventilation (Drager Primus, Germany) during the
the measured position of the bronchoscope as reported by the
procedure. Heartbeat and SpO2 were monitored continuously
on-chip sensor. The worst-case RMS registration error between
using pulse oximetry (Mindray Biomedical Electronics, Shen-
the two sets of data points is measured to be 5.8 mm. This error
zhen, China) during the procedure with an acceptable range
value represents the cumulative accuracy of both the magnetic
of 50 to 120 bpm and 90% to 100% SpO2 .
navigation system and the CT-to-patient registration error,
considering real-world factors such as breathing movement.
A. Endoscopic Targeting Accuracy Results To examine the effects of breathing motion on the accuracy
Five skin-based markers were used to assist in registration of the navigation, the sensor was held in a static position
to a pre-operative CT image (Siemens Somatom Definition on the chest of the animal, and its displacement over a
AS). A 128-slice CT with 0.6 mm slice thickness and 0.7 mm 100-second window was recorded. This data was recorded
slice pitch was used for segmentation with CT tube settings in 1-second increments and filtered using a 5-point moving
of 35 mA/120 kV. The l70f airway reconstruction kernel was average filter to make the underlying trend more apparent as
used for reconstruction followed by segmentation using Slicer shown in Fig. 18(c). Breathing-induced artifacts clearly reduce
5.2.2. Segmentation was achieved using Mimics 3D Medical the overall accuracy of both the registration and navigation.
Image Processing Software (Mimics 20.0, Materialise bv, In commercial adaption, it is possible to synchronize the
Leuven, Belgium) and visualised using Slicer. A customised breathing motion with the navigation system to minimize these
Slicer module using a modified iterative closest point (ICP) effects.
approach [72] which evaluates multiple initial conditions to These results highlight that the movement of the animal
avoid local minima was implemented for registration of the rather than sensor accuracy is the dominant source of naviga-
CT images to the animal anatomy for airway navigation. A tion error in this experiment. At the end of the investigation,
therapeutic bronchoscope (Olympus BF 1T1100) with a 3 mm the animal is euthanized under anesthesia with an intravenous
working channel was used for mapping and visualisation. For injection of pentobarbital 40 mg/kg IV (Euthoxin® , Osalia).
registration, the IC probe was placed in the working channel Table V provides a comparison of the performance achieved
of the bronchoscope and a balanced survey of the major left in this study with the state-of-the-art approaches [52], [53],
and right airways was completed using standard procedures [73] along with a commercially available sensor (610099,

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
14 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

Northern Digital Inc., Waterloo, Canada) for EMT [29], [74]. R EFERENCES
A notable advantage of this design is the significant reduction
[1] S. J. Spaner et al., “A brief history of endoscopy, laparoscopy, and la-
in overall sensor area, to 1.06 mm2 , which is of paramount paroscopic surgery,” Journal of Laparoendoscopic & Advanced Surgical
importance for minimally invasive image-guided interventions. Techniques, vol. 7, no. 6, pp. 369–373, 1997. PMID: 9449087.
[2] L. Boni et al., “Infective complications in laparoscopic surgery,” Surgical
B. Study Limitations Infections, vol. 7, no. supplement 2, pp. s–109–s–111, 2006.
[3] M. Kroh et al., “First human surgery with a novel single-port robotic
The current study details promising but preliminary results system: Cholecystectomy using the da vinci single-site platform,” Sur-
with the IC. One limitation of the study was the lack of gical endoscopy, vol. 25, pp. 3566–73, 06 2011.
[4] S. Wren and M. Curet, “Single-port robotic cholecystectomy results from
distortion correction at the sensor. To minimise the risk of a first human use clinical study of the new da vinci single-site surgical
distortion, care was taken to avoid sources of distortion such platform,” Archives of surgery (Chicago, Ill. : 1960), vol. 146, pp. 1122–
as metal beds or stainless steel tools. However, the device 7, 06 2011.
[5] C. for Devices, U. F. Radiological Health, and D. Administration, “Ini-
performed well inside the therapeutic bronchoscope and in the tiative to reduce unnecessary radiation exposure from medical imaging,”
presence of endoscopic instruments such as forceps. Feb. 2010.
The study targeting accuracy to registration was also clearly [6] D. Brenner, “Should we be concerned about the rapid increase in ct
usage?,” Reviews on environmental health, vol. 25, pp. 63–8, 03 2011.
limited by the segmentation accuracy and resolution of CT.
[7] A. Sulieman, G. Paroutoglou, A. Kapsoritakis, A. Kapatenakis,
These have no relevance to the IC sensitivity and, as is S. Potamianos, M. Vlychou, and K. Theodorou, “Reduction of radiation
typically the case in clinical practice, the overall accuracy of doses to patients and staff during endoscopic retrograde cholangiopan-
EM tracking is determined by the registration error rather than creatography.,” Saudi journal of gastroenterology : official journal of
the Saudi Gastroenterology Association, vol. 17, pp. 23–29, Feb. 2011.
the accuracy of the underlying hardware. The breathing motion Place: India.
captured by the chest-based on-chip sensor shown in Fig. 18(c) [8] L. Naidu, S. Singhal, D. Preece, A. Vohrah, and D. Loft, “Radiation
highlights this. exposure to personnel performing endoscopic retrograde cholangiopan-
creatography,” Postgraduate medical journal, vol. 81, pp. 660–2, 11
2005.
VI. C ONCLUSIONS [9] A. Roguin, J. Goldstein, and O. Bar, “Brain tumours among interven-
Electromagnetic tracking technology is playing a crucial tional cardiologists: A cause for alarm? report of four new cases from
two cities and a review of the literature,” EuroIntervention : journal of
role in complex medical procedures by reducing the need for EuroPCR in collaboration with the Working Group on Interventional
radiation-based imaging. However, it has not been widely used Cardiology of the European Society of Cardiology, vol. 7, pp. 1081–6,
in more common image-guided procedures like laparoscopic 12 2011.
[10] G. Mastrangelo, U. Fedeli, E. Fadda, A. Giovanazzi, L. Scoizzato,
surgery and non-robotic endoscopy due to expensive sensors, and B. Saia, “Increased cancer risk among surgeons in an orthopaedic
ranging from approximately $25 for 5-DoF devices to around hospital,” Occupational Medicine, vol. 55, pp. 498–500, 09 2005.
$250 for 6-DoF devices. This paper presents a cost-effective [11] D. Kaplan, J. Patel, F. Liporace, and R. Yoon, “Intraoperative radiation
safety in orthopaedics: A review of the alara (as low as reasonably
solution that utilizes compact 0.5 mm wide and 2.3 mm achievable) principle,” Patient Safety in Surgery, vol. 10, p. 27, 12 2016.
long on-chip magnetic sensors for 5-DoF magnetic tracking [12] G. Turchetti, F. Pierotti, and A. Cuschieri, “Economic evaluation of da
in image-guided interventions, making it both affordable and vinci-assisted robotic surgery: a systematic review,” Surgical endoscopy,
practical for real-world applications. The proposed on-chip vol. 26, pp. 598–606, 03 2012.
[13] A. Advincula and A. Song, “The role of robotic surgery in gynecology,”
sensor’s small size and cost-effectiveness enable its seamless Current opinion in obstetrics gynecology, vol. 19, pp. 331–6, 09 2007.
integration into existing magnetic navigation systems without [14] S. L. Chang, A. S. Kibel, J. D. Brooks, and B. I. Chung, “The impact
requiring significant modifications, while substantially reduc- of robotic surgery on the surgical management of prostate cancer in the
USA.,” BJU international, vol. 115, pp. 929–936, June 2015. Place:
ing the scalable cost to an estimated $1.50 compared to England.
approximately $25 for existing discrete 5-DoF sensors. Addi- [15] Z. Sun, A. Abaziz, and A. Yusof, “Radiation-induced noncancer risks
tionally, the on-chip sensor offers a digital readout, enhancing in interventional cardiology: Optimisation of procedures and staff and
patient dose reduction,” BioMed research international, vol. 2013,
the sensor’s robustness compared to the wire-wound sensor p. 976962, 01 2013.
with an analog readout. [16] A. Durán, S. K. Hian, D. L. Miller, J. Le Heron, R. Padovani, and
Furthermore, this work also presents a 6-DoF system that E. Vano, “Recommendations for occupational radiation protection in
interventional cardiology.,” Catheterization and cardiovascular inter-
combines low-cost on-chip sensors with wire-wound sensors ventions : official journal of the Society for Cardiac Angiography &
featuring ferromagnetic core devices. This system offers a Interventions, vol. 82, pp. 29–42, July 2013. Place: United States.
cost-effective, compact, and competitive solution in terms of [17] E. Vano, N. J. Kleiman, A. Duran, M. M. Rehani, D. Echeverri,
and M. Cabrera, “Radiation cataract risk in interventional cardiology
its form factor. The 6-DoF sensor achieves accuracy with personnel.,” Radiation research, vol. 174, pp. 490–495, Oct. 2010. Place:
a localization error of 0.8 mm and an angular error of 1.1 United States.
degrees over a 15×15×15 cm3 volume in real-time using an [18] K. Cleary and T. M. Peters, “Image-guided interventions: Technology
review and clinical applications,” Annual Review of Biomedical Engi-
AC magnetic field. The prototype sensor successfully demon- neering, vol. 12, no. 1, pp. 119–142, 2010. PMID: 20415592.
strated its ability to accurately track positions in in-vivo and a [19] A. Sorriento, M. B. Porfido, S. Mazzoleni, G. Calvosa, M. Tenucci,
worst-case registration accuracy of 5.8 mm was reported, albeit G. Ciuti, and P. Dario, “Optical and electromagnetic tracking systems
due mostly to patient motion artifacts rather than tracking in- for biomedical applications: A critical review on potentialities and
limitations,” IEEE Reviews in Biomedical Engineering, vol. 13, pp. 212–
accuracies. Overall, this technology holds significant promise 232, 2020.
for cost-effective, accurate, compact, and seamless integration [20] R. Newton et al., “Progress toward optical biopsy: Bringing the micro-
into current systems for magnetic tracking in image-guided or scope to the patient,” Lung, vol. 189, pp. 111–9, 02 2011.
[21] D. Nuernberg et al., “Efsumb recommendations for gastrointestinal ultra-
robotics interventions, offering new possibilities for improving sound part 3: Endorectal, endoanal and perineal ultrasound,” Ultrasound
patient care with cost-effective technology solutions. International Open, vol. 05, pp. E34–E51, 02 2019.

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 15

TABLE V
C OMPARISON WITH THE STATE - OF - THE - ART
Metrics This work [52] [53] [73] [55] [54] NDI Aurora [29] [74]
CMOS technology [nm] 65 65 65 65 No 180 130 No
Sensing Dimension 1D 2D 3D 3D N/A 3D 1D 1D
Localisation Dimension 3D (5-DoF) 3D(6-DoF) 3D 3D 3D 3D 2Da 3D
Localisation Accuracy [mm] 1.1 0.8 0.1b ⩽1 10 ⩽1 ⩽5 1.1
Angular Accuracy [Deg] N/A 1.1 N/A N/A N/A N/A N/A 0.07
Position update rate [Hz] 20 20 7 25 N.R. N.R. N.R. 15
Field of View [cm3 ] 15×15×15 15×15×15 20×20×10 4×8×12 38×27×24 40×40×40 ⩽5cm 25×25×25
Sensor On-chip On-chip + Discrete AK09970Nc On-chip LSM303Dd Passive HMC1052L Discrete
Total area [Chip] [mm2 ] 1.06e 1.06e 1.5 4 N/A 1.4 4 3.6f
Avg. Power [mW] 6g 6g 1 0.014 NA 0.336 0.912 NA
Magnetic Excitation Coil Size 8.5×8.5 cm2 8.5×8.5 cm2 30×30 × 1 cm3 N.R. N/A 10×10 cm2 N.R. N.R.
Current per Excitation Board (mA) 150 150 18000/X-Coil N.R. N/A. 32 N.R. N.R.
a b c d e
Only 1D tracking plot is shown Data reported from a fixed reference point on each coil in 1D Off-chip digital 3D hall effect sensor Off-chip tri-axes digital magnetic sensor on-chip sensor included
f g
only sensor area Power included all blocks shown in Fig. 2 N/A : Not applicable N.R.: Not reported

[22] F. T. W. Groenland et al., “Intravascular ultrasound-guided versus arrhythmias, and cardiac electrophysiology : journal of the working
coronary angiography-guided percutaneous coronary intervention in groups on cardiac pacing, arrhythmias, and cardiac cellular electro-
patients with acute myocardial infarction: A systematic review and meta- physiology of the European Society of Cardiology, vol. 19, pp. 1700–
analysis.,” International journal of cardiology, vol. 353, pp. 35–42, Apr. 1709, Oct. 2017. Place: England.
2022. Place: Netherlands. [38] K. O’Donoghue, H. A. Jaeger, and P. C. Murphy, “Sensor fusion
[23] S. andersson engels and P. Andersen, “Special section guest editorial: hardware platform for robust electromagnetic navigation,” in 2022 IEEE
Special section on selected topics in biophotonics: Optogenetics and International Instrumentation and Measurement Technology Conference
label-free optical spectroscopy,” Journal of Biomedical Optics, vol. 23, (I2MTC), pp. 1–6, 2022.
p. 1, 07 2018. [39] R. Cormier and Y. Bouslimani, “Electromagnetic and inertial motion
[24] N. Hardy, J. Dalli, M. F. Khan, P. Andrejevic, P. Neary, and R. Cahill, sensor fusion,” in 2021 IEEE International Symposium on Robotic and
“Inter-user variation in the interpretation of near infrared perfusion imag- Sensors Environments (ROSE), pp. 1–7, 2021.
ing using indocyanine green in colorectal surgery,” Surgical Endoscopy, [40] Y. Zhang, K. Wang, J. Jiang, and Q. Tan, “Research on intraoperative
vol. 35, pp. 1–8, 12 2021. organ motion tracking method based on fusion of inertial and electro-
[25] X. Jiang, J. Ma, G. Xiao, Z. Shao, and X. Guo, “A review of multi- magnetic navigation,” IEEE Access, vol. 9, pp. 49069–49081, 2021.
modal image matching: Methods and applications,” Information Fusion, [41] M. Cavaliere, D. Crowley, H. A. Jaeger, K. O’Donoghue, and
vol. 73, pp. 22–71, 2021. P. Cantillon-Murphy, “Magnetic model calibration and distortion com-
[26] K.-H. Englmeier and M. D. Seemann, “Multimodal virtual bronchoscopy pensation for electromagnetic tracking in a clinical environment,” IEEE
using pet/ct images,” Computer Aided Surgery, vol. 13, no. 2, pp. 106– Transactions on Magnetics, vol. 59, no. 7, pp. 1–12, 2023.
113, 2008. [42] H. Sadjadi, K. Hashtrudi-Zaad, and G. Fichtinger, “Simultaneous elec-
[27] B. L. van der Hoeven, M. J. Schalij, and V. Delgado, “Multimodal- tromagnetic tracking and calibration for dynamic field distortion com-
ity imaging in interventional cardiology,” Nature Reviews Cardiology, pensation,” IEEE Transactions on Biomedical Engineering, vol. 63,
vol. 9, pp. 333–346, June 2012. no. 8, pp. 1771–1781, 2016.
[28] A. M. Franz, T. Haidegger, W. Birkfellner, K. Cleary, T. M. Peters, [43] P. Sommer et al., “Initial experience in ablation of typical atrial flutter
and L. Maier-Hein, “Electromagnetic tracking in medicine—a review of using a novel three-dimensional catheter tracking system.,” Apr. 2013.
technology, validation, and applications,” IEEE Transactions on Medical ISSN: 1532-2092 1099-5129, Issue = 4 Journal Abbreviation: Europace
Imaging, vol. 33, no. 8, pp. 1702–1725, 2014. Pages: 578-581 Publication Title: Europace : European pacing, arrhyth-
[29] H. A. Jaeger, A. M. Franz, K. O’Donoghue, A. Seitel, F. Trauzettel, mias, and cardiac electrophysiology : journal of the working groups on
L. Maier-Hein, and P. Cantillon-Murphy, “Anser EMT: the first open- cardiac pacing, arrhythmias, and cardiac cellular electrophysiology of
source electromagnetic tracking platform for image-guided interven- the European Society of Cardiology.
tions.,” International journal of computer assisted radiology and [44] K. Klinge, O. Guntinas-Lichius, A. Naumann, and A. H. Mueller,
surgery, vol. 12, pp. 1059–1067, June 2017. Place: Germany. “Ultrasonography-guided electromagnetic needle tracking in laryngeal
[30] M. Cavaliere, O. McVeigh, H. A. Jaeger, S. Hinds, K. O’Donoghue, electromyography.,” European archives of oto-rhino-laryngology : offi-
and P. Cantillon-Murphy, “Inductive sensor design for electromagnetic cial journal of the European Federation of Oto-Rhino-Laryngological
tracking in image guided interventions,” IEEE Sensors Journal, vol. 20, Societies (EUFOS) : affiliated with the German Society for Oto-Rhino-
no. 15, pp. 8623–8630, 2020. Laryngology - Head and Neck Surgery, vol. 276, pp. 1109–1115, Apr.
[31] “Aurora sensors.” https://www.ndigital.com/electromagnetic-tracking- 2019. Place: Germany.
technology/aurora/aurora-sensors. [45] R. H. Taylor, N. Simaan, A. Menciassi, and G.-Z. Yang, “Surgical
[32] Z. Yaniv, E. Wilson, D. Lindisch, and K. Cleary, “Electromagnetic robotics and computer-integrated interventional medicine [scanning the
tracking in the clinical environment,” Medical physics, vol. 36, pp. 876– issue],” Proceedings of the IEEE, vol. 110, no. 7, pp. 823–834, 2022.
92, 04 2009. [46] T. Dammad and B. A. Jalil, Robotic Assisted Bronchoscopy, pp. 453–
[33] P. J. Reynisson et al., “Navigated Bronchoscopy: A Technical Review,” 463. Cham: Springer International Publishing, 2023.
Journal of Bronchology & Interventional Pulmonology, vol. 21, no. 3, [47] E. Folch, A. Mittal, and C. Oberg, “Robotic bronchoscopy and future di-
2014. rections of interventional pulmonology,” Current Opinion in Pulmonary
[34] E. E. Folch et al., “Electromagnetic Navigation Bronchoscopy for Medicine, vol. 28, no. 1, 2022.
Peripheral Pulmonary Lesions: One-Year Results of the Prospective, [48] X. Liu et al., “Preclinical evaluation of ultrasound-augmented needle
Multicenter NAVIGATE Study.,” Journal of thoracic oncology : official navigation for laparoscopic liver ablation.,” International journal of
publication of the International Association for the Study of Lung computer assisted radiology and surgery, vol. 15, pp. 803–810, May
Cancer, vol. 14, pp. 445–458, Mar. 2019. Place: United States. 2020. Place: Germany.
[35] M. Lu, S. Nath, and R. W. Semaan, “A Review of Robotic-Assisted [49] A. Romaszko and A. Doboszynska, “Multiple primary lung cancer: A
Bronchoscopy Platforms in the Sampling of Peripheral Pulmonary literature review,” Advances in clinical and experimental medicine :
Lesions.,” Journal of clinical medicine, vol. 10, Dec. 2021. Place: official organ Wroclaw Medical University, vol. 27, 04 2018.
Switzerland. [50] J. Jiang, S. Chang, A. Kent, T. Geraci, and R. Cerfolio, “Current novel
[36] S. Rolf et al., “Catheter ablation of atrial fibrillation supported by novel advances in bronchoscopy,” Frontiers in Surgery, vol. 7, 11 2020.
nonfluoroscopic 4D navigation technology.,” Heart rhythm, vol. 10, [51] J. D. Duke and J. Reisenauer, “Robotic bronchoscopy: potential in
pp. 1293–1300, Sept. 2013. Place: United States. diagnosing and treating lung cancer.,” Expert review of respiratory
[37] C. Piorkowski et al., “Mapping-guided characterization of mechanical medicine, vol. 17, pp. 213–221, Mar. 2023. Place: England.
and electrical activation patterns in patients with normal systolic function [52] S. Sharma, A. Telikicherla, G. Ding, F. Aghlmand, A. H. Talkhooncheh,
using a sensor-based tracking technology.,” Europace : European pacing, M. G. Shapiro, and A. Emami, “Wireless 3D Surgical Navigation

Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
16 GENERIC COLORIZED JOURNAL, VOL. XX, NO. XX, XXXX 2017

and Tracking System With 100µm Accuracy Using Magnetic-Field [74] L. Maier-Hein, A. M. Franz, W. Birkfellner, J. Hummel, I. Gergel,
Gradient-Based Localization,” IEEE Transactions on Medical Imaging, I. Wegner, and H.-P. Meinzer, “Standardized assessment of new elec-
vol. 40, pp. 2066–2079, Aug. 2021. tromagnetic field generators in an interventional radiology setting:
[53] S. Sharma, H. Melton, L. Edmonds, O. Addington, M. Shapiro, and Assessment of EM field generators,” Medical Physics, vol. 39, pp. 3424–
A. Emami, “A monolithic 3d magnetic sensor in 65nm cmos with ≤ 3434, May 2012.
10 µTrms noise and 14.8µW power,” in 2023 IEEE Custom Integrated
Circuits Conference (CICC), pp. 1–2, 2023.
[54] W.-F. Loke, W.-H. Chen, T. Maleki, M. Abu Khater, B. Ziaie, L. Papiez,
and B. Jung, “A 0.5-v sub-mw wireless magnetic tracking transponder
for radiation therapy,” in 2011 Symposium on VLSI Circuits - Digest of
Technical Papers, pp. 172–173, 2011.
[55] M. Rustom and C. Sideris, “Wireless frequency-division multiplexed
3d magnetic localization for low power sub-mm precision capsule en-
doscopy,” in 2022 IEEE Custom Integrated Circuits Conference (CICC),
pp. 01–02, 2022.
[56] H. A. Jaeger et al., “Peripheral tumour targeting using open-source
virtual bronchoscopy with electromagnetic tracking: a multi-user pre-
clinical study,” Minimally Invasive Therapy & Allied Technologies,
vol. 28, pp. 1–10, 11 2018. Manish Srivastava (M’21) is a Ph.D. student in
[57] M. Srivastava et al., “Electromagnetic tracking - on-chip real-time the School of Electrical and Electronic Engineer-
tracking for surgery,” Dec. 2022. U.K. Patent 2218643.1 filed on. ing at the University College Cork. Prior to this,
[58] J. Burghartz and B. Rejaei, “On the design of rf spiral inductors on he worked as an Analog circuit design engineer
silicon,” IEEE Transactions on Electron Devices, vol. 50, no. 3, pp. 718– at Synopsys INC from 2018 to 2019, and for
729, 2003. Qualcomm from 2016 to 2018. He has five is-
[59] A. Sidun, M. Srivastava, K. O’Donoghue, H. A. Jaeger, M. Cavaliere, sued patents and six pending patents. Manish’s
D. O’Hare, C. van den Bosch, and P. Cantillon-Murphy, “Planar on- current research focuses on the development of
silicon inductor design for electromagnetic tracking,” IEEE Sensors electromagnetic sensors and low noise sensor
Journal, vol. 23, no. 18, pp. 21129–21136, 2023. interface electronics circuits (such as the Low
[60] B. Razavi, Fundamentals of Microelectronics. Wiley, 2021. Noise CCIA and continuous time sigma delta
[61] H. Chandrakumar and D. Marković, “5.5 a 2µw 40mvpp linear- ADC) for use in navigation applications such as bronchoscopy, electro-
input-range chopper- stabilized bio-signal amplifier with boosted input physiology mapping, and biopsies, among others.
impedance of 300 MΩ and electrode-offset filtering,” in 2016 IEEE Kilian O’Donoghue received B.E. and PhD de-
International Solid-State Circuits Conference (ISSCC), pp. 96–97, 2016. grees in electrical and electronic engineering
[62] H. Jiang,
√ K. A. A. Makinwa, and S. Nihtianov, “9.8 an energy-efficient from University College Cork, Cork, Ireland, in
3.7nV/ Hz bridge-readout ic with a stable bridge offset compensation 2011 and 2015 respectively. He has over ten
scheme,” in 2017 IEEE International Solid-State Circuits Conference years’ of experience in medical electronic de-
(ISSCC), pp. 172–173, 2017. sign and has worked in multiple start-up and
[63] M. Maruyama, S. Taguchi, M. Yamanoue, and K. Iizuka, “An analog early-stage medical device companies in Ireland
front-end for a multifunction sensor employing a weak-inversion biasing and Canada, developing core technologies in
technique with 26 nvrms, 25 acrms, and 19 farms input-referred noise,” robotics, sensing, navigation, and medical imag-
IEEE Journal of Solid-State Circuits, vol. 51, no. 10, pp. 2252–2261, ing systems. His research interests include EM
2016. tracking systems, circuit design, magnetic field
[64] H. Jiang, modeling, and minimally invasive surgeries.
√ S. Nihtianov, and K. A. A. Makinwa, “An energy-efficient
3.7-nv/ hz bridge readout ic with a stable bridge offset compensation Aleksandr Sidun (BSc’18 MSc’20) received
scheme,” IEEE Journal of Solid-State Circuits, vol. 54, no. 3, pp. 856– the BSc degree in Electronic Engineering from
864, 2019. Saint-Petersburg Polytechnic University, Rus-
[65] Q. Fan, F.√Sebastiano, J. H. Huijsing, and K. A. A. Makinwa, “A 1.8 µ sia, in 2018 and completed his MSc at Saint-
w 60 nv/ hz capacitively-coupled chopper instrumentation amplifier Petersburg Polytechnic University in 2020. He
in 65 nm cmos for wireless sensor nodes,” IEEE Journal of Solid-State joined the Department of Electronic Engineer-
Circuits, vol. 46, no. 7, pp. 1534–1543, 2011. ing at Polytechnic University in 2018 designing
[66] S. Billa, S. Dixit, and S. Pavan, “Analysis and Design of an Audio analog front-ends for MEMS sensors. From 2020
Continuous-Time 1-X FIR-MASH Delta–Sigma Modulator,” IEEE Jour- to 2021 he was with Digital Solutions, Moscow
nal of Solid-State Circuits, vol. 55, pp. 2649–2659, Oct. 2020. where he designed high-speed ADCs and power
[67] S. Mondal, O. Ghadami, and D. A. Hall, “10.2 A 139 µ W 104.8dB- management circuits. In 2021 he joined the Mi-
DR 24kHz-BW CT∆ΣM with Chopped AC-Coupled OTA-Stacking and croelectronic Circuits Centre Ireland (MCCI) based at the Tyndall Na-
FIR DACs,” in 2021 IEEE International Solid- State Circuits Conference tional Institute, University College Cork designing power management
(ISSCC), (San Francisco, CA, USA), pp. 166–168, IEEE, Feb. 2021. circuits for electromagnetic tracking systems. He is currently an Analog
[68] A. Khashaba, J. Zhu, N. Pal, M. G. Ahmed, and P. K. Hanumolu, “A Design Engineer at Analog Devices International in Limerick designing
32-mhz, 34-w temperature-compensated rc oscillator using pulse density DC-DC converter circuits.
modulated resistors,” IEEE Journal of Solid-State Circuits, vol. 57, no. 5, Herman Alexander Jaeger is a lecturer in
pp. 1470–1479, 2022. the School of Engineering in University College
[69] B. Razavi, Design of CMOS Phase-Locked Loops: From Circuit Level Cork. He obtained his B.E, M.Eng.Sc and Ph.D
to Architecture Level. Cambridge University Press, 2020. degrees from University College Cork in 2014,
[70] L. Shen, N. Lu, and N. Sun, “A 1v 0.25 µw inverter-stacking amplifier 2015 and 2018 respectively. In 2015 he worked
with 1.07 noise efficiency factor,” in 2017 Symposium on VLSI Circuits, in Imperial College London at the Hamlyn Insti-
pp. C140–C141, 2017. tute for Robotic Surgery developing novel cam-
[71] B. K. P. Horn, H. M. Hilden, and S. Negahdaripour, “Closed-form era tracking algorithms for minimally invasive
solution of absolute orientation using orthonormal matrices,” J. Opt. surgery. From 2016 to 2017, Alex was a visiting
Soc. Am. A, vol. 5, pp. 1127–1135, Jul 1988. researcher in the Institute for Image Guided
[72] K. S. Arun, T. S. Huang, and S. D. Blostein, “Least-squares fitting of Surgery (IHU) in Strasbourg where he devel-
two 3-d point sets,” IEEE Transactions on Pattern Analysis and Machine oped Anser EMT, the world’s first open-source magnetic tracking system
Intelligence, vol. PAMI-9, no. 5, pp. 698–700, 1987. for guided surgical interventions. Following completion of his PhD in
[73] G. Shao, Y. Tang, L. Tang, Q. Dai, and Y.-X. Guo, “A novel passive 2018 he worked as a post-doctoral researcher at the Tyndall National
magnetic localization wearable system for wireless capsule endoscopy,” Institute from 2019 to 2022 where he worked on novel magnetic sensors
IEEE Sensors Journal, vol. 19, no. 9, pp. 3462–3472, 2019. for surgical navigation. In 2022 Alex joined the School of Engineering
in UCC as a lecturer in electrical and electronic engineering where he
teaches courses in introductory power systems and non-linear circuit
Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024
analysis. In 2023 athe20:34:05
was awardedUTC from
anIEEE Xplore.
SFI-IRC Restrictions
Pathway award apply.
to further
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
his electromagnetic tracking research in VR/AR applications.
 This article has been accepted for publication in IEEE Transactions on Biomedical Circuits and Systems. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TBCAS.2024.3384016
AUTHOR et al.: PREPARATION OF PAPERS FOR IEEE TRANSACTIONS AND JOURNALS (FEBRUARY 2017) 17

Alessandro Ferro graduated with a Bachelor’s Pádraig Cantillon-Murphy is a Professor of

Degree in Information Engineering from the Uni- Biomedical Engineering at University College
versity of Padua, Italy, in 2018. He is pursuing Cork, Ireland, academic member of Tyndall Na-
a Master of Engineering Science (MEngSc) in tional Institute at University College Cork, and an
Electrical and Electronic Engineering at Univer- honorary faculty at l’Institut de Chirurgie Guidée
sity College Cork (UCC), Ireland. Since 2021, par l’Image in Strasbourg. He graduated with a
he has been a member of the Microelectronic first-class honours B.E. degree (2003) in Elec-
Circuits Centre Ireland (MCCI) at the Tyndall Na- trical and Electronic Engineering from Univer-
tional Institute, contributing to the development sity College Cork, Ireland before completing his
of on-chip sensing to track surgical instruments Master of Science (2005) and Ph.D. (2008) de-
for the Anser EMT project. He is an active mem- grees at the Department of Electrical Engineer-
ber of the IEEE Cork student branch, focusing his research interests ing and Computer Science at Massachusetts Institute of Technology
on integrating technology in medical devices to enhance surgical proce- (MIT). His current research interests include electromagnetic tracking
dures. and navigation in image-guided interventions, and surgical robotics.
His teaching interests include electronic circuits, electromagnetics, and
biomedical design. He is a former Marie Curie fellow (2010-2014), a
Daragh Crowley (Member, IEEE) received the former MIT Whitaker fellow (2007-08), a senior member of the IEEE, and
B.E. degree in electrical and electronic engi- an ERC awardee (2020). He has co-founded two start-up companies.
neering from University College Cork (UCC),
Cork, Ireland, in 2021. He is currently pursu-
ing the Ph.D. degree in electrical and elec-
tronic engineering with the Tyndall National In-
stitute, Cork. His current research interests in-
clude electromagnetic tracking, embedded sys-
tems and image-guided surgical navigation.

Christian van den Bosch (Senior Software

Engineer, Deep Field project) received his B.Sc.
(Hons.) degree in computer science from Uni-
versity College Cork (UCC), Cork, Ireland, in
2007. Their current research interests include
electromagnetic tracking (EMT), robotics, and
the prototyping and characterisation of medical
micro-coils for EMT.

Professor Marcus Kennedy (MD FRCPI

FCCP) is currently a respiratory physician with
interest in interventional pulmonology (IP) at
Cork University Hospital and Honorary Pro-
fessor of Medicine at University College Cork
(UCC). He is the clinical lead of the Respiratory
Department at Cork UNIVERSITY Hospital and
currently President of the Irish Thoracic Society.
He was the chair of the National Cancer Control
Program lung cancer group from 2017-2020 and
chair of the first National Lung Cancer Guideline.
He has over 80 peer reviewed publications including NEJM, Thorax,
CHEST and AJRCCM publications and authored numerous book chap-
ters. He has been a PI or co-investigator on multiple grants. His current
research interests includes trials in single use bronchoscopes, remote
simulation training in bronchoscopy and electromagnetic navigation
tracking.

Daniel O’Hare (S’13 M’07) received the BE

degree in Electronic Engineering from University
College Dublin, Ireland, in 2000 and completed
his PhD at the University of Limerick in 2017. He
joined Motorola Semiconductor 2000 and from
2004 to 2008 he was with Freescale Semicon-
ductor designing ADCs and DACs for Cellular
transceivers. From 2008 to 2012 he was with
M4S NV a spinout of IMEC and from 2013 to
2017 he was an ADC researcher in the Circuits
and Systems group at the University of Limerick.
He is currently a Principal Researcher in the Microelectronic Circuits
Centre Ireland (MCCI) based at the Tyndall National Institute, University
College Cork. His research is in low-noise sensor interface circuits
and integrated circuits to enable medical systems. This includes the
design of low voltage, low noise, data-converters, and analog signal
conditioning circuits. Dr. O’Hare was co-recipient of the 2021 NEWCAS
best student paper. He has more than 15 published papers and 2
granted Patents.
Authorized licensed use limited to: UNIVERSITY COLLEGE CORK. Downloaded on May 07,2024 at 20:34:05 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.