Magnetic Resonance Elastography in the Liver at 3 Tesla
Using a Second Harmonic Approach
D.A. Herzka,1,2* M.S. Kotys,1,3 R. Sinkus,4 R.I. Pettigrew,5 and A.M. Gharib5
Magnetic resonance elastography (MRE) using mechanical
stimulation has demonstrated diagnostic value and clinical
promise in breast, liver, and kidney at 1.5 Tesla (T). However,
MRE at 1.5T suffers from long imaging times and would benefit
from greater signal-to-noise for more robust postprocessing.
We present an MRE sequence modified for liver imaging at 3.0T.
To avoid artifacts in the phase images, the sequence maintains
a short TE by using a second harmonic approach, including
stronger motion encoding gradients, shorter radio frequency
pulses and an echo-planar readout. Scan time was decreased
by a factor of
2 relative to 1.5T by using an EPI readout and a
higher density sampling of the phase waveform was used to
calculate shear stiffness and viscosity. Localized (small region
of interest) and global (whole-liver region of interest) measure-
ments in normal healthy subjects compared very favorably with
previously published results at 1.5T. There was no significant
difference between global and localized measures. Magn Re-
son Med 62:284–291, 2009. © 2009 Wiley-Liss, Inc.
Key words: liver fibrosis; magnetic resonance elastography;
elastography; liver stiffness
Magnetic resonance elastography (MRE) using mechanical
stimulation has demonstrated diagnostic value in the clin-
ical areas of breast and liver imaging examinations at 1.5
Tesla (1–6). Liver MR elastography has been used to mea-
sure liver stiffness in the progression to and staging of
hepatic fibrosis. Given the clinical promise of these tech-
niques, mechanically stimulated elastography has become
a desirable objective for regular clinical use. However,
classic three-dimensional (3D) MRE using spin echo or
gradient echo sequences at 1.5T suffers from long imaging
times on the order of 20 min (2) and would benefit from
greater signal-to-noise for more robust postprocessing.
Therefore, a transition to 3.0T may be the next logical step
in making MRE more practical in the routine clinical set-
ting because the higher field strength brings the promise of
higher intrinsic signal-to-noise ratio (SNR) (7). However,
transition to a higher field also brings important concerns,
such as greater susceptibility and technical challenges.
Most notably, previous studies at 1.5T have used a longer
echo time (TE) (8) that becomes untenable at 3.0T due to
signal loss.
Here, we take advantage of the higher intrinsic SNR of
the MR signal phase at 3.0T to implement several technical
changes for high-field MRE. First, we use second harmonic
imaging, which is similar to fractional encoding of har-
monic motion (9), and permits image acquisitions with
shorter TEs. Second, we apply a shorter radio frequency
(RF) pulse (fewer lobed-sinc) under the assumption that
the initial 90-degree pulse will provide the necessary se-
lectivity. Third, we exploit stronger gradient magnitudes
(maximum amplitude of 60 mT/m) to compensate for the
loss of sensitivity associated with second harmonic imag-
ing. Finally, we increase the echo train length to reduce
overall scan time (10). A 3.0T liver MRE sequence was
demonstrated and tested in normal human volunteers. Re-
sults in the form of shear stiffness and shear viscosity
values are compared with previously published results
from MR elastography at 1.5T.
THEORY
Imaging at 3.0T requires a short TE to minimize signal loss
arising from enhanced sensitivity to susceptibility and
motion artifacts. Moreover, T2 in the liver is shorter at
3.0T. The bipolar gradients needed for motion encoding
are time-consuming when synchronous with the low fre-
quency of mechanical stimulation essential for sufficient
penetration of mechanical waves. For example, the me-
chanical frequency (m) of 65 Hz used in previous studies
at 1.5T (1) extends the TE to 61 ms as it requires two
15.4-ms-long bipolar encoding gradients. An example
MRE pulse sequence using a mechanical frequency of
60 Hz is shown in Figure 1. By using motion sensitizing
gradients (MSGs) that oscillate at the second harmonic of
m, the time required for one complete gradient waveform
period can be reduced by half while remaining sensitive to
the main driving frequency m of the mechanical stimulus.
Certainly, this leads to a reduced sensitivity to small dis-
placements and is apparent from the calculation of the
phase accumulation  due to the mechanical standing
wave sinmt induced in tissue:

A


0
TE
Gtsinmt ddt [1]
where  is the gyromagnetic ratio of protons, A is the peak
amplitude of the mechanical wave, G(t) represents the
applied MSGs in the measurement, phase, or slice direc-
tion, and m 2
m. The mechanical wave can be mea-
sured at multiple points on the displacement curve by
1National Heart, Lung and Blood Institute, NIH, DHHS, Bethesda, Maryland.
2Clinical Sites Research Program, Philips Research North America, Bethesda,
Maryland.
3MR Clinical Science, Philips Healthcare, Cleveland, Ohio.
4Laboratoire Ondes et Acoustique, ESPCI, Paris, France.
5National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda,
Maryland.
Drs. Herzka and Kotys contributed equally to this work.
*Correspondence to: Daniel Herzka, Department of Biomedical Engineering,
Johns Hopkins University School of Medicine, 720 Rutland Avenue, 726 Ross
Bldg., Baltimore MD 21205. daniel.herzka@gmail.com
Received 7 April 2008; revised 9 December 2008; accepted 18 December
2008.
DOI 10.1002/mrm.21956
Published online 15 May 2009 in Wiley InterScience (www.interscience.wiley.
com).
Magnetic Resonance in Medicine 62:284–291 (2009)
© 2009 Wiley-Liss, Inc. 284

adding a time delay (tdel) which produces a phase shift
(d m
tdel). G(t) can be expressed as follows:
Gt


G
singt for t
t1 . . . t2 and t3 . . . t4
0 otherwise [2]
where g 2
g, g is the frequency of the sinusoidal
MSG, and the time points t1,t2,t3,t4 define the start and
finish of the gradients before and after the refocusing pulse
(as shown on Fig. 1). G represents the maximum amplitude
Time (ms)
Readout
Phase
Slice
RF
0 20 60 30 40 50 10
TE
3.0 Tesla - 2nd Harmonic
t1
t2 t3
t4
tdel
Transducer
60Hz
Transducer
Readout
Phase
Slice
RF
0 20 60 30 40 50 10
Time (ms)
60Hz
TE
1.5 Tesla - 1st Harmonic
t1 t2 t3 t4
tdel
a
b
FIG. 1. MR elastography spin echo pulse sequence diagrams for both 1.5T and 3.0T. a: At 1.5T, the sinusoidal motion sensitizing gradients
(MSGs, blue) oscillate at the same frequency as the mechanical transducer (60 Hz in this example). b: For 3.0T, the MSGs oscillate at twice the
frequency, reducing the minimum TE achievable. A shorter refocusing RF pulse further reduces TE while the higher amplitude MSGs available with
stronger gradient hardware recover some of the lost motion encoding. Finally, three-echo EPI is used to increase the SNR efficiency of the
sequence at 3.0T. Note that the MSGs are applied in only one of the three encoding directions during each of the three scans, but are shown in
all three for display purposes only. The complete image acquisition is repeated eight times, with different tdel to sample the sinusoidal phase in
a robust manner. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]
3T Liver MRE Using a Second Harmonic Approach 285