Magn Reson Mater Phy (2007) 20:273–278
DOI 10.1007/s10334-007-0098-7
RESEARCH ARTICLE
Magnetic resonance elastography compared with rotational
rheometry for in vitro brain tissue viscoelasticity measurement
Jonathan Vappou · Elodie Breton ·
Philippe Choquet · Christian Goetz ·
Rémy Willinger · André Constantinesco
Received: 12 September 2007 / Revised: 18 November 2007 / Accepted: 29 November 2007 / Published online: 15 December 2007
© ESMRMB 2007
Abstract Magnetic resonance elastography (MRE) is an
increasingly used method for non-invasive determination of
tissue stiffness. MRE has shown its ability to measure in
vivo elasticity or viscoelasticity depending on the chosen
rheological model. However, few data exist on quantitative
comparison of MRE with reference mechanical measurement
techniques. MRE has only been validated on soft homoge-
neous gels under both Hookean elasticity and linear visco-
elasticity assumptions, but comparison studies are lacking
concerning viscoelastic properties of complex heterogeneous
tissues. In this context, the present study aims at compar-
ing an MRE-based method combined with a wave equation
inversion algorithm to rotational rheometry. For this purpose,
experiments are performed on in vitro porcine brain tissue.
The dynamic behavior of shear storage (G ′) and loss (G ′′)
moduli obtained by both rheometry and MRE at different
frequency ranges is similar to that of linear viscoelastic prop-
erties of brain tissue found in other studies. This continuity
between rheometry and MRE results consolidates the quan-
titative nature of values found by MRE in terms of viscoelas-
tic parameters of soft heterogeneous tissues. Based on these
results, the limits of MRE in terms of frequency range are
also discussed.
Keywords Magnetic resonance elastography · Linear
viscoelasticity · Phase encoding · Brain biomechanics
J. Vappou · R. Willinger
Institut de mécanique des fluides et des solides,
UMR 7507 CNRS-Université Louis Pasteur,
Strasbourg, France
e-mail: jvappou@imfs.u-strasbg.fr
E. Breton · P. Choquet · C. Goetz · A. Constantinesco (
B
)
Service de biophysique et médecine nucléaire,
CHRU Hautepierre, Strasbourg, France
e-mail: andre.constantinesco@chru-strasbourg.fr
Introduction
Magnetic resonance elastography (MRE) is a fast growing
research field as well as a technique that is being increas-
ingly used to determine tissue stiffness. After the theoret-
ical approach proposed by Lewa [1], MRE was originally
described in 1995 [2] as a palpation-like diagnostic tool for
detection of abnormal stiffness differences in soft tissues.
Under an external mechanical harmonic excitation, and by
means of a motion-synchronized encoding gradient included
in the MRI sequence, it is possible to observe acoustic shear
waves propagating inside the sample. The displacement map
resulting from this wave propagation allows one to determine
the rheological properties of the medium. For several years,
and also in many recent studies, MRE has been based on the
assumption of shear waves propagating in a purely Hookean
elastic medium [2–7]. Despite this limiting assumption, MRE
has been shown to be an efficient tool for diagnostic purposes
where an estimation of a qualitative global stiffness value is
usually sufficient [4–6].
More recently, inversion reconstruction techniques have
been developed to determine quantitative viscoelastic param-
eters under assumption of linear viscoelasticity. This has led
to the further development of this method, turning MRE into
a powerful tool for the characterization of linear viscoelas-
tic properties of soft tissues. In particular, such experiments
have been performed on breast [8] and brain [9] tissue, pro-
viding a particularly novel in vivo distribution measurement
of viscoelastic properties of the concerned organs.
However, few data exist in the literature that allow for a
comparison between the results obtained by MRE and those
obtained with conventional mechanical techniques. MRE has
only been compared on soft homogeneous tissue-mimicking
gels. Such comparisons have either been performed assuming
Hookean elasticity (with compression–extension tests [2,10]
123

274 Magn Reson Mater Phy (2007) 20:273–278
and with dynamic oscillatory tests [11]) or assuming linear
viscoelasticity (with dynamic oscillatory measurements
[12]). A comparison is missing between MRE and other
mechanical techniques when investigating the viscoelastic
properties of complex organs.
In this context, the present study offers to determine the
dynamic viscoelastic properties of in vitro brain tissue by
MRE in order to compare them with rotational rheometry
performed on small brain samples.
Material and methods
Both rheometric and MR experiments were performed on
normal porcine brains (aged from 6 to 8 months) obtained
from a slaughterhouse and immediately refrigerated at 4◦C.
Total post-mortem time varied between 24 to 48 hours. Exper-
iments were performed on eight brains, four for each method,
under the same experimental conditions (post-mortem time
range and conservation) until the beginning of the tests.
For rheometric experiments, cylindrical-shaped samples
(20 mm in diameter, 4–5 mm high) were excised from white
matter in the corona radiata region. Samples were placed in a
parallel plate geometry and sand paper was glued to lower and
upper plates of a rotational rheometer (AR2000, TA-instru-
ments, DE, USA) in order to limit possible slippage between
plates and sample. A moist chamber was used in order to pre-
vent dehydration and experiments were performed at 25◦C.
A total of 12 samples were tested at ε = 0.5% strain in
the 0.1–10 Hz frequency range. Raw phase was systemati-
cally checked in order to ensure appropriate inertia correc-
tion, and critical value of frequency limit ( f = 10 Hz) was
determined.
MR experiments were performed on a low-field 0.1 T resis-
tive magnet (Bouhnik S.A.S., Velizy-Villacoublay, France)
using a modified 2D spin-echo sequence. Two cycles of
square-shaped motion encoding gradients (magnitude≈
15 mT/m) were included into the phase encoding direction.
MRI and excitation devices are detailed in [12] and illus-
trated in Fig. 1. MR sequence used in this study is repre-
sented in Fig. 2. Typical MR parameters were TR = 800 ms,
TE = 60 ms, field of view FOV = 90 × 90 mm, slice thick-
ness 10 mm, NEX = 8, matrix size 128 × 96 (reconstructed
128×128), total acquisition time of 10 min for one image.
Excitation frequency range was 80–140 Hz with 20 Hz steps
and amplitude of harmonic motion was approximately of
10µm as measured by accelerometry prior to experiments.
The porcine brains were placed in a cylindrical container
and were plunged into echographic gel in order to increase
RF coil loading [and therefore to enhance signal to noise
ratio (S/N)], to settle mechanically the pig brain and to pre-
vent its dehydration. A rectangular actuator (3 mm thick ×
80 mm long × 30 mm wide) covered with sand paper was
Fig. 1 Experimental device (left) showing magnet, exciter, and
magnetic field B0 orientation. Schematic view (right) or location of
actuator between left (L) and right (R) hemispheres of brain
Fig. 2 MRI sequence pulse chronogram used in this study, show-
ing synchronization between the motion-encoding gradient and the
mechanical wave, the RF line, and the slice (S), phase (P), and read
(R) encoding directions
placed between the two hemispheres of the brain, in mid
sagittal plane. Experiments were realized at a temperature of
27◦C ± 2◦C.
An inverse wave equation inversion 2D-algorithm was
used for determination of real and imaginary parts of com-
plex wave number k from temporal Fourier transforms of
displacement. Similar inverse approaches have been used by
several authors for determination of viscoelastic properties
by MRE [8,9,13]. For each frequency, eight phase offsets
between motion and motion-sensitizing gradient were used,
leading to a total acquisition time of 80 min. The compo-
nent of wavefield used for reconstruction was the direction
parallel to motion of the actuator (phase direction). Phase
images were processed by application of a 2D Gaussian
filter (5 × 5 convolution matrix and standard deviation of
2 pixels) as a low-pass frequency filter. Incompressibility
of brain tissue was supposed, neglecting therefore the
123