MIT OpenCourseWare

http://ocw.mit.edu

HST.583 Functional Magnetic Resonance Imaging: Data Acquisition and Analysis

Fall 2008

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
HST.583: Functional Magnetic Resonance Imaging: Data Acquisition and Analysis, Fall 2008
Harvard-MIT Division of Health Sciences and Technology
Course Director: Dr. Randy Gollub.

HST.583
October 3, 2008
Problem Set #2

Problem 1: Understanding K­space

Describe what you would see AND explain why when you look at a T1 image of the mid­
sagittal slice through the brain after:
a) discarding the LOW frequency k­space information
b) discarding the HIGH frequency k­space information
You may include images, but a verbal description is required.

a) After discarding the low frequency k­space information, the resulting image
would have the same detail of the original image, but would have a lower signal level.
This is because the high frequency region which is retained encodes the finer details of
the signal, but the low frequency region which is discarded is responsible for the higher
signal­to­noise ratio of the image.
b) After discarding the high frequency k­space information, the resulting image
will have a similar signal to the original image, but will have a lower spatial resolution
due to the loss of the high frequency information which encodes the finer details of the
image.

Problem 2: Factors that affect the spin level populations

The difference in level population in a two level system is denoted by Δ n = Nα – Nβ,
where N is the number of spins and the levels are denoted by α or β. The time rate of
change of Δ n is given by d(Δ n)/dt = ­(Δ n ­ Δ n0)/T1, where T1 is the spin­lattice
relaxation time and Δ n0 is the population difference at equilibrium.
a) Find the equation for the population difference at some time, t, after placing the
sample in the magnet?
b) How long does it take for the distribution to get to 90% of the Boltzmann
distribution if the T1 is 800 ms?

Page 1 of 7
We want Δ n(t) and we have d(Δ n)/dt = ­(Δ n ­ Δ n0)/T1.

Rearranging,

d(Δ n) / (Δ n ­ Δ n0) = ­dt/T1

Integrating both sides and recalling that integral(dx/x) = ln(x),

ln(Δ n ­ Δ n0) = ­t/T1 + C, where C is a constant arising from indefinite integration

limits. Since C is constant, we can set t=0 and arrive at

C = ln(Δ n(0) ­ Δ n0)

ln(Δ n ­ Δ n0) ­ ln(Δ n(0) ­ Δ n0) = ­t/T1

Recall that ln(x) – ln(y) = ln(x/y)

ln[ (Δ n ­ Δ n0) / (Δ n(0) ­ Δ n0) ] = ­t/T1

Raise both sides by e and recall that eln(x) = x

(Δ n ­ Δ n0) / (Δ n(0) ­ Δ n0) = e­t/T1

Δ n(0) is the spin level difference of the sample immediately after being placed in the
magnet, and cannot change instantaneously. Therefore, Δ n(0) = 0, the value it had
before being placed in the magnet.

Δ n(t) = ­ Δ n0e­t/T1 + Δ n0 = Δ n0(1­ e­t/T1)

b) We want Δ n(t) = 0.9Δ n0

0.9Δ n0 = Δ n0(1­e­t/T1)

0.1 = e­t/T1

Take the natural log of both sides and recall that ln(ex) = x

t = ­T1*ln(0.1) = 1.84 sec

Problem 3: Playing with T1 and T2 contrast

Page 2 of 7
Using the table of tissue parameters below, we will investigate how the choice of pulse
sequence parameters can change the contrast in the image. We will assume that the pulse
sequence is the following: [π /2 – t1 ­ π /2 – (t1+t2) ­ π /2 – t1 – acquisition], a
stimulated echo, and make use of the relationship: S = (ρ /2)( exp –(t1+t2)/T1 ) exp ­2t1/T2,
where S is the MR signal strength, ρ is the proton spin density, t1=TE/2 and t1+t2=TM.

Tissue T1 T2 ρ
Gray matter 1.2 s 70 ms .98
White matter 800 ms 45 ms .80

a) Plot the signal equation versus t1 for both white matter and gray matter assuming
t2 = 50 ms.

b) In another figure, plot the signal difference between white matter and gray matter
as t2 is changed. Assume t1 = 25 ms.

Page 3 of 7
Problem 3: Calculation of Imaging Gradients
Assuming a 10 cm field of view, a 10 ms phase encoding gradient, and 8 phase encoding
steps, what are the phase angles for each phase encoding step?

θ y = γ Δ y Gy τ
Gy: FOVy = 1/γ Gy τ  Gy = 1/γ FOVy τ
θ y = γ Δ y τ (1/γ FOVy τ )
θ y = Δ y / FOVy
θ y = 1 cm / 10 cm = .1 rad

Problem 4: Human safety in MRI

The goal of this problem is to make you aware of human safety issues when using fast
brain imaging sequences. Consider the following two pulse sequences: single shot
gradient­echo EPI (i.e., one 90° RF excitation pulse followed by a train of gradient
recalled echoes), and a single shot spin­echo sequence (i.e., one 90° RF excitation pulse
followed by a train of 180° RF refocusing pulses that form spin­echoes).
a) Find an expression for the specific absorption rate (SAR) for both sequences.
Hint: make a simple model that considers the RF power deposited for a single RF
pulse and weight it by the duty cycle for the train of RF pulses. The duty cycle is
the amount of time the RF is turned on per TR, thus it contains information about
each RF pulse duration, the number of slices acquired and the number of RF
pulses per TR. Does the weight of the patient matters?
b) Based on the expression found in the previous point, compare the two sequences
and discuss their limitations in terms of SAR. Which of the two sequences would
be likely to induce more peripheral nerve stimulation?

Page 4 of 7
Without consideration of the orientation of the object to be imaged, the SAR will be
proportional to: integral(B12(t)*dt) * [duty cycle] / weight(kg), where B1(t) is the
applied RF pulse, the duty cycle is the the number of RF pulses per TR (this takes into
account the number of slices acquired per TR), and the limits of integration are 0 to TR.
Since there is only 1 RF pulse per slice for the gradient echo sequence and several for the
spin­echo sequence, the spin­echo sequence will distribute more energy into the subject
per unit time given the same TR. For part b), the crucial factor is how fast the gradient
fields change, i.e. fast gradient field switching is usually responsible for peripheral nerve
stimulation. Therefore, the gradient echo sequence would be most likely to cause PNS.

Problem 5: Considerations in pulse sequence acquisition parameters

Which parameter(s) would you keep the same to ensure that the degree and direction of
distortion between two different pulse sequences were the same? Why?

Assuming an EPI sequence and imaging the same volume (or slice), the main factors that
affect distortions (where distortions does not include blurry images from only sampling
the center of k­space) are bandwidth and echo spacing for the degree of the distortion,
and the readout direction for the orientation of the distortion. For example, if we switch
the frequency encoding and phase encoding directions, the ghosts you mentioned will be
along the orthogonal dimension to what they were in the original sequence.

Page 5 of 7
Problem 6: More on Pulse Sequence Acquisition Parameters
For each of the two pulse sequences shown below:
a) Calculate the Total Acquisition Time
b) Calculate the Voxel Size
c) State whether it is a structural or functional scan (and how you know)
d) State whether it is a 2D or 3D scan

SEQUENCE ONE
Scanner SIEMENS MAGNETOM TrioTim
Routine
Slab group: 1
SEQUENCE TWO
Orientation: Sagittal Scanner SIEMENS MAGNETOM TrioTim
Phase encode direction: A >> P Routine
Readout direction: R >> L Slice group 1
Slices per slab: 160 Slices: 30
FoV read: 220 mm Dist. Factor: 25 %
FoV phase: 220 mm Orientation: Transversal
Slice thickness: 1.20 mm Phase encode direction: A >> P
TR: 2300 ms Read encode direction: R >> L
TE: 2.94 ms FoV read: 220 mm
Averages: 1 FoV phase encode: 220 mm
Contrast Slice thickness: 4.0 mm
Magnetization preparation: Non­sel. IR TR: 2000 ms
TI: 1100 ms TE: 30 ms
Flip angle: 9 deg Averages: 1
Reconstruction: Magnitude Contrast
Measurements: 1 MTC: Off
Resolution Flip angle: 77 deg
Base resolution: 256 Measurements: 142
Phase resolution: 192 Resolution
Slice resolution: 256 Base resolution: 64
PAT mode: GRAPPA Phase resolution: 64
Acceleration factor: PE 2 PAT mode: None
Geometry Matrix Coil Mode: Triple
Multi­slice mode: Single shot Geometry
Series: Interleaved Multi­slice mode: Interleaved
Sequence Series: Ascending
Bandwidth: 240 Hz/Px Sequence
Echo spacing: 7 ms Bandwidth: 2298 Hz/Px
RF spoiling: On Free echo spacing: Off
Echo spacing: 0.5 ms
EPI factor: 64
Dummy Scans: 3

Page 6 of 7
Scan1 Scan 2
a). The total imaging time is the number a). This is an EPI sequence used for
of phase encodes multiplied by the TR fMRI acquisition. 30 slices with the
time. Since we have GRAPPA (parallel given in­plane resolution are acquired
imaging technique) with acceleration within the rather long TR period. The
factor of 2, the scan time should be total acquisition time would depend on
almost halved. So we have 192*2.3s/2 = the number of measurement, the TR time
200.8 sec = 3.68min. Note that the and the dummy scans. Therefore, we
GRAPPA algorithm requires acquiring have (142+3)*2 = 290sec = 4.833min.
several ACS (or phase­encoding) lines
for proper reconstruction, which add to b). The slice thickness is 4mm and in
the scan time. The lab handout does not plane matrix size is 64x64 over FOV is
have this info, so I can’t say what would 220mm2. Therefore the voxel size is
be the precise imaging time, but in 4x(220/64)x(220/64)mm3 = 47.27mm3 =
reality it should be a bit more than the 0.0473cc.
calculated 3.68min. Further, all the 160
slices are acquired within the one TR, c). This is a functional acquisition since
and this is possible since the TR is long the number of measurements is 142 and
and the TE is short. the TE and TR are long, and the
resolution is low.
b). The FOV in x and y is 220mm, and
the matrix sizes are 256 and 192 d). This is a multislice 2D scan.
respectively. Therefore, for x and y, we
have (220/256)x(220/192) =
0.86x1.15mm2. Since the slice thickness
is 1.2mm the overall voxel size is
0.86x1.15x1.2mm3 = 1.1868mm3.

c). This is a structural acquisition since

the number of measurements is 1.

d). We are acquiring 160 slices, so this is

a 3D scan.

Page 7 of 7