MRI:

image formation and sequences

Medical devices and systems

Maurizio Schmid
2017-18
 Image formation: resonance
– Let’s consider slice selection has been performed: according to what has
been introduced in the previous lesson, an elementary volume placed at
position (x,y) will respond to the gradient pulses according to the equation
below:
ds(x,y,t) = exp(jφ(y))*exp(jω(x)t)*ρ(x,y)dxdy
– Thus representing the fact that along one direction, the position is
encoded though the phase, and along the other, it is encoded through the
frequency.
– This can be also expressed in the following alternative way, by
considering the overall response of all the elements inside the volume
7 6
Sτpe (t)=∫(7 ∫(6 𝜌 𝑥, 𝑦 𝑒 ()*+,-./ 𝑒 ()*+01 2304 𝑑𝑥𝑑𝑦
– For one specific τpe we will have a signal function of t; by repeating for
different values of τpe we will have a 2-d function of both t and τpe.
 Image formation: resonance
7 6
S(t,τpe)=∫(7 ∫(6 𝜌 𝑥, 𝑦 𝑒 ()*+,-./ 𝑒 ()*+01 2304 𝑑𝑥𝑑𝑦
– If we call kx = ɣGphτpe and ky = ɣGfrt, we have:
7 6
S(t,τpe)=∫(7 ∫(6 𝜌 𝑥, 𝑦 𝑒 ()89. 𝑒 ()8:2 𝑑𝑥𝑑𝑦
– This means that, by considering the 2d-matrix obtained juxtaposing along
one direction the MRI RF response to each RF pulse obtained with a
specific τpe duration, we can calculate ρ(x,y) as the 2-D Fourier transform
of this obtained 2-d function.
– In reality, the part that is not related to the localization does not contain
ρ(x,y) only, but it will also include the relaxation part.
– The transformation in the spatial frequency domain is also called
transformation in the k-space.
Image formation: procedure
• Thus, 1-D data
obtained for a specific
τpe are then Fourier
transformed along the
two directions, to
obtain the
representation of
ρ(x,y).
• Multiple phase
responses are then
needed to obtain a 2d
image for a specific
slice..
 Image as a 2-D Fourier transform of the RF response

To the image

From the RF
responses
 t à ky
τpe à kx
t à ky x

y x
Image as a 2-D Fourier transform of the RF response
2D Fourier from k-space: example
2D Fourier from k-space: example
 High and low frequencies in the k-space
K-space Image

Complete data

Reconstruction from LF only

(squared area). Smoothing, loss
in details,

Reconstruction from HF only

(squared areas). Info only on
details, discontinuities (along
the corresponding directions).
 Back to the Free Induction Decay
• The FID captured by the RF coil will thus be proportional to :
• FID(x,y)=(M0sina)*cos(Wyt+Fx)*exp(-t/T2*)
• This includes localization characteristics, and tissue component
characteristics:
• While T2 (or, more in general, T2*) is directly linked to the FID through
the relaxation part, T1 does not directly appear in the equation.
• In order to make the FID be dependent from T1 we will use the a
angle, through the so called excitation sequences. Excitation
sequences are also used to lose dependency of time constant T2 from
field inhomogeneity (T2*).
 Recall: T1 and T2
• T1 represents the time constant with which the vertical magnetization goes back to
the original, when only B0 is applied.
• T2 represents the time constant with which the transverse magnetization goes
back to 0, when only B0 is applied.
 Recall: T1 and T2
• Thus, two differen relaxation processes:
T1: lattice-spin relaxation.

T2: spin-spin relaxation.

FID: single pulse excitation
 Excitation sequences: summary
• Sequences are composed of appropriately lagged in time RF
pulse sequences.
• Recall that the receiver RF coils are able to just capture the
transverse magnetic field (no longitudinal component).
• Each pulse i-th of a sequence will be denoted by two
parameters:
– the “intensity” αi, corresponding to the amount of flip angle α
that the pulse is able to transfer t the magnetization
– the delay ti, corresponding to the time delay from the previous
pulse.
 Excitation sequences: summary
• Each sequence will last a predefined time TR.
• This TR is made of the sum of the different delays
between each pulse sequence.
• A sequence is then denoted by juxtaposing the
characteristic of each pulse of the sequence, with the
following notation:
• (a1-t1-a2-t2………am-tm)n
• where:
– ai is an RF pulse able to force an ai degrees flip angle
– ti is the time interval between the i-th and the i+1-th pulse
– Siti = TR ;
– n represents the number of times the sequence needs to be repeated.
 Excitation sequences
• The number of possible excitation sequences
is very high: we will focus on the three main
ones, that are/were most popular:

– Saturation recovery
– Spin echo
– Inversion recovery
• STIR
• FLAIR
– Turbo sequences
 Saturation Recovery - SR
• Objective:
– To make it possible to obtain T1-weighted images.

• Procedure:
– Repeated application of 90° pulses, with time distance TR.
Sequence (90°-TR)n
 Saturation Recovery
90o 90° RF single pulse.

Mz
Mxy

FID

With a single pulse, Mz goes back to M0. But what if a successive 90° pulse is
applied before going back to normal (for instance, TR = 2T1)?
 Saturation Recovery
90o
TR<T1

Mz
Mxy TR>>T1

Mxy2

FID FID2
In general, the second FID (Mxy) will have an amplitude proportional to:
é æ TR öù
r ê1 - expçç - ÷÷ú
êë è T1 øúû
If TR<T1, this relation is heavily dependent from T1. By repeating this multiple times, the
successive FID maxima will thus follow the spin-lattice relaxation time T1. It is thus a T1-
weighted imaging sequence. If instead TR > T1, this will lead to a fid dependent from ρ.
 Tissue difference

No difference

TR

TR

By appropriately choosing TR to distinguish between tissues, once T1 values are known, we

can maximize contrast according to differences in T1.
 Spin Echo - SE
• Objective:
– To make it possible to obtain both T1-weighted and T2-weighted
images. Remove dependence from field inhomogeneity.

• Procedure
– Repeated application of a 90° RF pulse, followed, after a time equal to
TE/2 by an 180° RF pulse.
• Sequence: (90°-TE/2-180°-TE)n
 Spin Echo
• After applying the 90° RF pulse, the pulse at 180° makes
it possible to “realign the spins”, thus generating an echo
response.
• Phase 1:

t=0 (90° RF pulse)

•Spins are aligned and grouped together

along one axis (in this case, x).
 Spin Echo
Phase 2: t ≤ TE/2 (before applying the 180° RF
pulse)
Spins start to fringe, i.e. some of them have a
higher frequency (V in the figure), some of them
are slower (L in the figure). This happens because
of the presence of field inhomogeneity (T2*).
Overall alignment is decreasing along the xy plane.
Phase 3: t ≥ TE/2 (after applying the 180° RF pulse)
This pulse overturns the positions of all the spins,
in such a way that the fastest spins are now lagging
behind, while the slowest ones are ahead. Still,
they continue to spin at their own frequency.
 Spin Echo

Medical devices and systems – Schmid – 2017-18

Spin Echo
Phase 4: t = TE

At this time instant, the spins basically realign,

showing a maximum for FID: but their
alignment is not perfect, since spin-spin
relaxation is maintained. Basically, the
difference in amplitude between t=0 and t=TE
depends only from T2.
Fringing and rephasing
 Spin Echo
90o 180o
TE/2 TE/2

T2* T2*

FID FID2
In the case of Spin Echo FID is proportional to:
é æ TR öù æ TE ö
ç
r ê1 - expç - ÷
÷ ú expçç - ÷÷
êë è T1 øúû è T2 ø
TR>>T1 and TE<<T2 à signal is dependent on ρ
TR<=T1 and TE<<T2 à signal is T1-weighted.
TR>>T1 and TE³T2 à signal is T2-weighted.
TE

Choosing a TE

Effect on contrast
 Spin Echo: examples
• By choosing different values of TR and TE we can obtain images with different
contrasts

T1-weighted, TR=14 ms, TE=5ms T2-weighted, TR=4000 ms, TE=100ms

T1 (ms) T2(ms)
B0=1.5T

TR >> T1 and TE << T2 à ρ-weighted CSF 4000 500

TR <= T1 and TE << T2 à T1-weighted
White matter 700 80-90
TR >> T1 and TE <= T2 à T2-weighted
Gray matter 1000 70-80
 Spin Echo: examples

Head, axial, 22 cm FOV, 5 mm slice thickness, SE

Left: [TR, TE] = [5500, 105] ms, 2 Nex, matrix 512x256
Right: [TR, TE] = [450, 14] ms, 1 Nex, matrix 256x192
(Nex: number of averaged images)
 Inversion Recovery - IR
• Objective:
– Enhance contrast.
• Procedure:
– First 180° RF pulse; then after a time Ti (inversion time), a 90° RF pulse,
and then signal read. (180°-Ti-90°)n
• The first pulse determines the inversion of the longitudinal magnetization.
• The second pulse let the receiver coil read the longitudinal remaining component
that has been partially recovered after Ti.
 Inversion Recovery
180o 90o

Mz

T2*

FID
Ti T
TR
FID read along the xy axis is proportional to:
é æ Ti ö æ T öù
r ê1 - 2 expçç - ÷÷ + expçç - R ÷÷ú
êë è T1 ø è T1 øúû
 Inversion Recovery: Ti selection
• The inversion time (the time
after which the 90° RF pulse
is applied after the 180° can
be chosen based on different
objectives.
• Tissues that have a shorter
T1 will contribute more than
those with longer T1.
• Repetition time TR needs to
be chosen in such a way that
original longitudinal
magnetization is recovered.
 TR

Ti < T1

TR

Ti ≈ T1

Ti determines contrast based on T1.

 Short Tau Inversion Recovery - STIR
• STIR: it’s an inversion recovery sequence in for which Ti is
chosen around T1fatln2. In tis way, fat response is
suppressed, since, for that value, the longitudinal
magnetization is zero.
• For instance, for a 1.5 T B0, T1 is approximately 150 ms.

Medical devices and systems – Schmid – 2017-18

 FLuid Attenuation Inversion Recovery - FLAIR
• FLAIR is an inversion recovery sequence conceptually similar to
STIR: in this case, T1 will be the one of the fluid one wants to
attenuate.
• It is used in head MRI, to attenuate response from Cerebro-Spinal
Fluid CSF. Very useful in the case of brain haemorrhage/bleeding.
CT MRI-FLAIR MRI-T2-w

Medical devices and systems – Schmid – 2017-18

 Turbo Sequences
• Turbo sequences have been proposed to decrease exam times.
• To do that, instead of waiting for the entire relaxation to be completed before
repeating the excitation, the repetition is done soon after the transverse
component is back to zero (and in some cases, field gradients are used to speed
up the process).
• In this way, we can reduce the time for each repetition, but we will have a
“smaller” signal, since the longitudinal component has not been totally recovered.
• Thus, more sensitive receivers are necessary.
• It is often used in MRI cardio exams.

https://www.youtube.com/watch?v=G4dFVeP9Vdo
Medical devices and systems – Schmid – 2017-18
 Contrast-based MRI
• Gadolinium is generally used (paramagnetic contrast
agent, 7 unpaired electrons). It is injected.
• It drastically decreases T1 of the tissues that are in
contact with it (blood) à very useful in case of bleeding,
but also inflammation (e.g. multiple sclerosis).
 Functional MRI (fMRI)
• MRI images are generally structural/anatomical, i.e. they
contain information on the anatomy of the tissues (and their
composition.
• Functional imaging refers instead to the techniques that are
used to estimate functions associated with those
tissues/organs.
• In MRI, it is possible to gain knowledge on functions from
images in different ways:
– Diffusion MRI (MRI imaging obtained to characterize diffusion of water
molecules)
– Perfusion MRI (Contrast agent to capture perfusion)
– Spectroscopy (Metabolic characteristics associated with protons)
– fMRI-BOLD (see next)
 fMRI-BOLD
• BOLD technique (Blood Oxygen Level Dependent),
which is now synonym for fMRI at the brain level:
– Uses an endogenous contrast agent - deoxygenated
haemoglobin;
– Leverages on the variations of oxygenation in the venae, as
the by-product of neural activation
• Active neurons need blood to provide them with energy to fire.
– Allows viewing the activation of brain regions, with
temporal resolution of a few seconds, and spatial
resolution around a few millimetres.
fMRI-BOLD
• Neural activation phases:
1. Neural activation
2. Increase of blood flow in the brain district
3. Increase of oxygenated haemoglobin (there
is an increase that is higher than what
needed)
4. Decrease of relative concentration of
deoxygenated haemoglobin.
5. T2* is sensitive to the decrease of
deoxygenated haemoglobin.
 fMRI
• Thus, the amount of neural
activity could be estimated by
considering changes in contrast
associated with one specific
activity, as compared to the
baseline.
• fMRI signal is thus obtained by
“subtracting” from the response
during an activity the one that is
obtained when no activity is
specified.
• Statistical analysis is crucial for
that.
– http://www.pnas.org/content/1
13/28/7900.full

Medical devices and systems – Schmid – 2017-18

 fMRI: protocols
• In order to determine the neural activation during an activity:
– The participant is subject to at least two different conditions (one or
multiple task, and a control one).
– EPI (Echo Planar Imaging) is performed;
– Statistical analysis is performed;
– “Active” voxels are then highlighted
fMRI - Neural Activation
 fMRI-BOLD and neural activation

• Logothetis, 2002: http://rstb.royalsocietypublishing.org/content/royptb/357/1424/1003.full.pdf

MRI test object
MRI test objects
 MRI: 3D reconstruction
• Brain neoplasms
• https://www.youtube.com/watch?v=3qnbA1TIh3E
• Virtual colonoscopy
• https://www.youtube.com/watch?v=lvnP8lt_5Qk
• Cardio MRI and blood flow
• https://www.youtube.com/watch?v=1ORpeB6j0gc
• https://www.youtube.com/watch?v=osgSmEMGSyk

Medical devices and systems – Schmid – 2017-18