dti of trabecular bone marrow
TRANSCRIPT
Magnetic Resonance Im
DTI of trabecular bone marrow
Cristina Rossia,b,d, Silvia Capuanic,d, Fabrizio Fasanod,e, Marcella Alesiania,b,d,
Bruno Maravigliaa,b,d,TaIstituto Nazionale Fisica della Materia, Centro di Ricerca e Sviluppo SOFT (INFM CRS-SOFT) c/o Universita’ di Roma bLa SapienzaQ, I-00185 Roma, Italy
bPhysics Department, University bLa Sapienza,Q 00185 Rome, ItalycINFM CRS SOFT
dMuseo Storico della Fisica e Centro Studi e Ricerche bEnrico Fermi Q, Compendio Viminale I-00184 Roma, ItalyeLaboratory of Functional Neuroimaging, Fondazione Santa Lucia, IRCCS, 00179 Rome, Italy
Abstract
The development of NMR diffusion imaging and diffusion tensor imaging (DTI) has offered the possibility of studying the porous
structures beyond anatomical imaging. In fact, random molecular motions, within tissue components, probe tissue microstructures.
Up to now, the DTI method was mainly used to investigate cerebral morphology and study white matter diseases. In this study, it has
been applied to trabecular bone marrow analysis to obtain structural information on spongy bone tissue. Our first results show that DTI
could represent an important tool in studying the microstructural architecture of the trabecular bone as well as the microarchitecture of
porous media.
D 2005 Elsevier Inc. All rights reserved.
Keywords: DWI; DTI; Trabecular bone; Anisotropic diffusion; Restricted diffusion
1. Introduction
Diffusion is the process by which matter moves across a
system as a result of random molecular motions. As Einstein
explained in a 1905 article, in a dilute solution solvent,
molecules hit solute particles continuously and from all
sides. This results in a random motion known as the
brownian motion [1]. The self-diffusion coefficient D [1]
provides a complete description of diffusion in homoge-
neous and isotropic systems. When molecular diffusion
occurs in heterogeneous and anisotropic media, the mea-
sured diffusion coefficient value depends on observation
direction and measurement duration.
By means of suitable NMR pulse sequences such as the
pulsed gradient spin echo (PGSE) [2], it is possible to
measure an apparent diffusion coefficient value. This
coefficient is not a true measure of the intrinsic diffusion
because it depends on the interactions of the diffusing
molecules with tissue microstructures [3]. Furthermore,
0730-725X/$ – see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.mri.2004.11.018
T Corresponding author. Physics Department, University bLa Sapi-
enza,Q Piazzale Aldo Moro 2, 00185 Rome, Italy. Tel.: +39 0649913473;
fax: +39 0649913484.
E-mail address: [email protected] (B. Maraviglia).
because diffusion is a three-dimensional process, a diffusion
tensor,D [1], rather than a scalar, D, is needed to describe the
mobility of brownian particles along the different directions.
The estimate of diffusion tensor components from MR
images enables us to create parametric images where pixel
intensity is proportional either to mean diffusivity (MD) or
to the degree of anisotropy of the media. The set of all MRI
techniques that allow to determine the elements of such a
tensor and display the information it contains in each voxel
is known as diffusion tensor imaging (DTI) [4].
Since the early 1990s, diffusion-weighted imaging
(DWI) has assumed an important role both in clinical
routine and in the research environment [4–7] because of
its high sensitivity to changes in the displacement
characteristics of water molecules in human tissues.
Because of this property, DWI represents a powerful tool
in the field of brain studies and brain disease diagnosis. As
an example, a promising application is due to the ability of
DWI to detect an early cerebral infarct with respect to other
conventional MR imaging techniques. Furthermore, in
cerebral tissues, white matter containing fibrous compo-
nents, where diffusion of water molecules is much more
restricted across than along the fibers results in anisotropic
diffusion (i.e., the diffusion coefficient is dependent on the
aging 23 (2005) 245–248
rf
diff
slice
phase
read
90° 90° 90°
TE/2 TM TE/2
t
∆δ
Fig. 1. Schematic rf pulse and gradient sequence for diffusion-weighted
stimulated echo imaging using three 908 rf pulses in conjunction with a
magnetic field gradient for slice selection (slice). The stimulated echo
signal, which occurs at TE+TM, is acquired in the presence of a frequency-
encoding gradient (read) and the whole sequence is repeated with different
phase-encoding gradients (phase). Diffusion weighting is obtained by a pair
of diffusion gradients (diff). D is the separation of the diffusion gradients of
duration d.
C. Rossi et al. / Magnetic Resonance Imaging 23 (2005) 245–248246
direction of the applied field gradient [3]), DTI [6] has also
made imaging the fiber tracts in the white matter of the
brain possible.
Up to now, DTI was mainly used to investigate cerebral
morphology [3]. In fact, outside the brain, the short T2 values
of the other body tissues force the use of shorter TE than the
ones used for brain tissue in order to minimize T2 weighting.
This involves the use of short diffusion gradient pulses
and requires hardware modification (such as gradient
strength and gradient shape) and ad hoc pulse sequence
[e.g., the pulsed gradient stimulated echo (PGSTE) se-
quence is better than the PGSE one] and parameters (such as
the diffusion time D and or the gradient pulse duration d)especially in media characterized by an high degree of
anisotropy and restricted diffusion such as porous media.
In this study, the DTI method was used for the first time to
obtain anisotropic information on trabecular bone tissue.
Parametric images representing the MD and the degree of
anisotropy [fractional anisotropy (FA)] were obtained sam-
pling an excised bovine spongy bone at 7 T. Our first results
show the great potentiality of DTI methods to study bone
tissue microstructure and tissue architecture arrangement.
Fig. 2. (A) MD map of a portion of celery. This map underlines the
differences in molecular diffusion regimes inside the sample. The bbulkQ ofthe vegetable (parenchyma) appears isointense, as we can expect for an
isotropic medium. On the other hand, the pixels that correspond to the
fibers appear less intense as a result of restricted diffusion regime. The MD
provides, in fact, a description of the overall diffusion in different sample
regions. The MD value is reported in mm2/s. (B) The FA map confirms the
goodness of the experimental protocol. In fact, the pixels that correspond to
the fibers are characterized by a great intensity, as we expect from an axial
geometry, while the map is darker in the parenchyma region, as we expect
from isotropic diffusion.
2. Methods and materials
The diffusion tensor D [8,9] is a symmetric tensor and its
elements are all positive, so it is possible to diagonalize it to
obtain the eigenvalues D1, D2 and D3 and the corresponding
eigenvectors k1, k2 and k3. The local frame of reference is
an orthogonal coordinate system with axes that are parallel
to the tensor eigenvectors. In such frame of reference, the
anisotropy can be visualized by means of the diffusion
ellipsoid, where D1, D2 and D3 are the principal diffusiv-
ities. The diffusion ellipsoid is a three-dimensional repre-
sentation of the diffusion distance covered in space by
molecules in a given diffusion time.
It is possible to define a set of scalar quantities that
enable visualization of the diffusion properties of the tissue.
Some of these quantities are invariant with respect to the
rotation of the coordinate system and, thus, are independent
from the laboratory frame in which the tensor components
are measured [9].
In this study, we considered the MD [9,10]:
MD ¼ D1þD2þD3
3ð1Þ
which corresponds to the average ellipsoid’s size and is
proportional to the tensor’s trace. It is a scalar invariant that
Fig. 3. T2-weighted image of an axial slice of a portion of bovine epiphysis
extracted by a bovine femur covered by a layer of fat. Signal from fat is
more intense due to its higher T2 value compared with spin–spin relaxation
time of marrow within the bony porous structure.
ig. 4. MD and FA maps of an axial slice of a portion of epiphysis covered
y a layer of fat. (A) The signal intensity of the MD map varies according to
e MD variation inside the fat and the spongy bone. The MD value is
ported in mm2/s. (B) The FA map confirms the highly isotropic molecular
iffusion inside the fat, where the pixels appear less intense. In the FA map,
e intensity of the pixels from the spongy bone is characterized by a great
ariability that suggests a variation in pore size and trabecular bone
rientation. As suggested by the comparison with the T2-weighted image
ported in Fig. 3, such structure-related features are only observable from
arametric maps obtained with DTI techniques.
C. Rossi et al. / Magnetic Resonance Imaging 23 (2005) 245–248 247
allows us to monitor the overall mean-squared displacement
of molecules, which is mainly determined by the presence
and distribution of obstacles to diffusion.
The FA [9,10]:
FA ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3½ðD1�MDÞ2þðD2�MDÞ2þðD3�MDÞ2�
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 D2
1þD22þD2
3
��q ð2Þ
measures the degree of anisotropy and is related to the
presence of oriented structures. Geometrically, it is propor-
tional to the ellipsoid’s eccentricity.
In our study, the PGSTE sequence [2] was used to obtain
a series of MR images from bovine spongy bone marrow.
The intensity of the gradient was varied for each of the
seven directions of the diffusion-encoding gradients to
collect a set of diffusion-weighted images along at least
six noncollinear gradient directions [11,12]. One unweight-
ed image (i.e., an image acquired without a diffusion-
encoding gradient) was also acquired to estimate the
components of D in each voxel and obtain the parametric
MD and FA images by means of a homemade software. All
images were obtained on the same slices (5-mm thickness).
The PGSTE sequence is especially suited for spin
systems characterized by T2bT1 [3]. It is therefore useful
to investigate restricted diffusion at different diffusion times.
In fact, in the PGSTE sequence, the diffusion time is limited
by the value of the mixing time, TM, during which the
longitudinal component of the magnetization decays with
the spin-lattice relaxation time, T1. In this work, the choice
of using a PGSTE sequence is related to the short T2 that
characterizes the sample. In fact, for fat’s hydrogen nuclei in
bovine bone sample, we obtain a T2 equal to about 31 ms
and a T1 equal to about 390 ms.
All the measurements were implemented and performed
on a 7 T Bruker Biospec horizontal magnet, equipped with
gradient unit with a maximum intensity of 280 mT/m, 300 Asrise time. An Avance digital spectrometer, with XWINNMR
and ParaVision 2.1 software version, was employed for data
acquisition and analysis. The experimental results were
obtained by using a microimaging probe (12 mm internal
diameter) for RF transmission and signal detection.
The experimental procedure was first tested on a fresh ce-
lery sample and then applied on a portion of extracted bovine
epiphysis excised from distal femur, covered by a layer of fat.
The trabecular bone consists of a three-dimensional
network where bone marrow is dispersed in the interstitial
F
b
th
re
d
th
v
o
re
p
C. Rossi et al. / Magnetic Resonance Imaging 23 (2005) 245–248248
spaces. The heterogeneous structure of spongy bone
exhibits simultaneously anatomical site dependence and
directional anisotropy of mechanical properties and archi-
tecture. In such a structure, marrow is trapped inside pores
of 50 Am–1 mm diameter, while the separation between
pores is about 200 Am: this geometry causes an aniso-
tropic molecule diffusion of the bone marrow in contrast
with the homogeneous structure represented by the fat
layer covering.
3. Results
The experimental procedure and our homemade software
that allowed reconstructing the MD and the FA maps were
tested and optimized on a standard phantom made of a fresh
celery portion. The conventional PGSTE sequence provided
with the appropriate slice, phase and read encoding gradients
for the imaging (Fig. 1) (d=4 ms, D=43 ms, TE=26.1 ms,
TR=6 s) was used, with the b values ranging from 0 to about
1550 s/mm2. The MD map in Fig. 2 confirms the isotropy of
molecular diffusion inside the vegetable inner part (paren-
chyma). On the other hand, the pixels corresponding to the
fibers appear less intense in brightness as a consequence of
the restricted molecular diffusion along the fiber axis. A
strong diffusion anisotropy inside the fibers (FA about 1)
and a large diffusion isotropy in the bulk of the sample (FA
about 0) were found in the FA map, in good agreement with
the expected trend.
The DWI sequence was then applied to obtain a series of
MR images from bovine bone samples. The sample was
extracted from the most extreme region of the spongy bone,
the epiphysis. The sample is constituted by a portion of
trabecular bone, filled with bone marrow and covered by a
layer of fat. So, both porous anisotropic system (trabecular
bone) and homogeneous isotropic system (fat) are present in
the sample. The T2-weighted image in Fig. 3 shows a great
difference between fat and bone marrow signals but does
not show any difference inside the trabecular structure.
To obtain the MD and FA maps shown in Fig. 4, d=4 ms
and D=220 ms were chosen for each gradient direction (x,
y, z, xy, yz, xz, xyz). Experimental b values ranged from 40
to 12000 s/mm2 (TR=5 s, TE=28 ms). The MD map in
Fig. 4A and the FA map in Fig. 4B show a great variability
in the intensity inside the trabecular structure. These results
can be explained by supposing a variation in pore size,
bigger near the fat and smaller at the bottom of the sample.
The FA image shows a typical artefact probably arising
from uncorrected eddy currents-induced distortions [13]. In
fact, the parametric image of the sample is characterized by
a hem of high anisotropy.
The absence of such an artefact in the celery FA map
may confirm our guess; in fact, we expect that the smaller
the diffusion gradient strength is, the better the eddy
currents compensation is and the smaller the artefact is.
The MD and FA images show different contrasts with
respect to conventional T2-weighted image, particularly in
the trabecular bone region. In particular, in epiphysis regions,
because of different contrasts in the FA image, different
microstructural rearrangement orientations are present. On
the other hand, the isotropy of diffusion in the fat region of
the sample is underlined by FA values around 0.
4. Conclusion
In this study, the DTI contrast between differently
structured areas of bovine spongy bone was investigated.
Because of bone marrow in the trabecular bone T2bT1,
PGSTE-based imaging sequences were used to obtain MD
and FA parametric images. Our preliminary results show
that MD and FA maps are more sensitive than conventional
NMR imaging techniques to bone trabecular network
microstructure. However, it is important to point out that,
in our experiments, the selected diffusion range of about
10 Am is less wide than the average dimension of the pores
in the trabecular bone (equal to 50–200 Am).
DTI proved to be a potential tool in studying bone
architecture and, in general, in assessing the geometrical
organization of porous systems. In fact, the MD map
underlines the difference in restricted diffusion (smaller or
bigger pores) while the FA map emphasizes the anisotropy
in a porous network.
References
[1] Crank J. The mathematics of diffusion. Oxford7 Clarendon Press;
1957.
[2] Callaghan PT. Principles of nuclear magnetic resonance microscopy.
Oxford7 Clarendon Press; 1991.
[3] Beaulieu C. The basis of anisotropic water diffusion in the nervous
system—a technical review. NMR Biomed 2002;15:435–55.
[4] Le Bihan D, editor. Diffusion and perfusion magnetic resonance
imaging. New York7 Raven Press Ltd; 1995.
[5] Moseley ME, Cohen Y, Mintorovitch J, Chileuitt L, Shimizu H,
Kucharczyk J, et al. Early detection of regional cerebral ischemia in
cats: comparison of diffusion- and T2-weighted MRI and spectros-
copy. Magn Reson Med 1990;14:330–46.
[6] Pierpaoli C, Jezzard P, Basser PJ, Barnett A, Di Chiro G. Diffusion
tensor MR imaging of the human brain. Radiology 1996;201:637–48.
[7] Warach S, Chien D, Li W, Ronthal M, Edelman RR. Fast magnetic
resonance diffusion-weighted imaging of acute human stroke.
Neurology 1992;42:1717–23.
[8] Basser PJ, Mattiello J, Le Bihan D. Estimation of the effective self-
diffusion tensor from the NMR spin-echo. J Magn Reson 1994;
103B:247–54.
[9] Basser PJ, Mattiello J, Le Bihan D. NMR diffusion tensor
spectroscopy and imaging. Biophys J 1994;66:259–67.
[10] Le Bihan D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N,
et al. Diffusion tensor imaging: Concepts and applications. J Magn
Reson Imaging 2001;13:534–46.
[11] Papadakis NG, Xing D, Houston GC, et al. A study of rotationally
invariant and symmetric indices of diffusion anisotropy. Magn Reson
Imaging 1999;17:881–92.
[12] Basser PJ, Pierpaoli C. A simplified method to measure the diffusion
tensor from seven MR images. Magn Reson Med 1998;39:928–34.
[13] Basser PJ, Jones DK. Diffusion-tensor MRI: Theory, experimental
design and data analysis—a technical review. NMR Biomed 2002;
15:456–67.