Transcript here - U-System - University of Arizona

Inferior Brain Lesions Monitored Using Diffusion Weighted Magnetic Resonance Imaging
Lars Ewell1, Naren Vijayakumar2, Jeffrey J. Rodriguez2 and Baldassarre Stea1
1) Department of Radiation Oncology, University of Arizona Medical Center, Tucson AZ 85724 USA 2) Department of Electrical and Computer Engineering, University of Arizona, Tucson AZ 85721 USA
Mutual Information
ABSTRACT
Radial Diffusion (RD) and Echo Planar (EP) Diffusion
Weighted Magnetic Resonance Image (DWMRI) scans of
five patients were compared in inferior brain locations.
Registration to non-diffusion T2 weighted FLuid
Attenuated Inversion Recovery (FLAIR) MRI scans was
used for comparison. Mutual Information (MI) was
utilized as a registration metric. RD DWMRI was, in
general, found to have a better image registration (higher
mutual information) to the T2 FLAIR images than EP
DWMRI. This advantage increased for more inferior
brain MRI slices.
Since the images of the human brain considered here are subject to
minimal deformation, a registration metric that utilizes rigid
transformations and/or rotations is appropriate. Mutual information
(MI) is the metric that has been chosen. For the axial scans in this
study, a three-dimensional Cartesian coordinate system was chosen,
such that the z axis is roughly parallel to the spine, with the x and y
axes along the left-right and anterior-posterior, respectively. This
choice is depicted
in Figure 1. When registering two MRI scans via MI, we are thus
searching for translations along the x and y axes, and rotations about
the z axis, such that the two images are as optimally aligned as
possible.
DWMRI Scans
The five image sets used for this study were obtained from an
internal review board (IRB) approved protocol. They consist of a
series of axial DWMRI scans taken from a superior position (top of
skull) to an inferior one (bottom of cerebrum or below). A sample of
both the EP and RD (DWMRI, b=0), as well as the T2 FLAIR images
to which they are registered can be seen in Figure 2.
Table 1: MI Between Different DWMRI Scans and T2 FLAIR
Scan Set
1
2
3
4
5
No. Slices
6
6
5
5
5
Mean EP MI
1.24
1.15
1.27
1.52
1.30
Mean RD MI
1.74
1.94
1.81
2.00
1.82
Max Diff. Loc
6
6
5
5
3
Introduction
Discussion
The use of DWMRI has increased recently, as the
application has spread from use mainly in ischemia diagnosis,
to monitoring therapy efficacy in radiation oncology1,2. When
considering isotropic diffusion, the most common form of
weighting is EP3. While this form has proven useful, lately
there have been additional novel forms that have been
proposed, such as RD4. In order to quantitatively compare
these two different forms of DWMRI, we have used image
registration to non-diffusion weighted scans, where mutual
information5 has been chosen as the metric.
Figure 2: Inferior (a, d, g), mid-brain (b, e, h) and
superior (c, f, i) MRI slices of RD, EP (DWMRI, b=0)
scans, and the T2 FLAIR images to which they are
registered, respectively. Note: DWMRI (EP and
RD) had window and level adjustments for clarity.
The mutual information between these primary images, and the
DWMRI scans was then calculated as in (3), as translations in the x
and y directions, as well as rotations about the z axis were varied. A
total of five patient scan sets were analyzed for this study.
Motivation
When treating brain tumors with radiotherapy, the lesion is
generally contoured by a clinician on a non-diffusion weighted
image, such as T2 FLAIR. In order to calculate an ADC for the
lesion, used to monitor therapy efficacy2, the contour must be
transcribed to a diffusion weighted image. Therefore, the ability
to utilize DWMRI to monitor therapy efficacy depends on the
accuracy with which these contours can be transcribed.
Although less common, gliomas can occur in more inferior
brain locations, such as the occipital lobe6. In positions near
magnetic inhomogeneities such as these and others, there are
well known shortcomings of single-shot EP DWMRI: geometric
distortions, blurring, poor spatial resolution and intra-voxel
dephasing. In view of this fact, there is a need for a more
robust form of DWMRI. With this in mind, we have compared
EP DWMRI with RD DWMRI.
A typical plot of the MI from the two different forms of DWMRI as a
function of slice number is shown in Figure 3. As can be seen in this
figure, the MI between the T2 FLAIR primary images and the RD
DWMRI scans is consistently higher than between T2 FLAIR and EP
DWMRI scans. In addition, this difference generally gets larger as the
axial slices go from a superior location to a more inferior one. The
differences between the MI of the different registrations are displayed
in Table 2. As can be seen in this table, the MI between the T2 FLAIR
and RD DWMRI scans is approximately 0.5 higher than the MI
between the T2 FLAIR and the EP DWMRI scans. In addition, the
maximal difference generally occurs in the most inferior scans (highest
slice number).
Figure 1: Coordinate system used
for mutual information registration.
As shown above, radial diffusion has a significantly higher value of mutual
information than echo planar when registered with a T2 FLAIR image
commonly used to contour brain lesions in preparation for radiation
treatment. In general, this difference becomes more pronounced as the
axial slice location in the brain moves from a superior position to an inferior
one.
Conclusion
The ability to clearly image a DWMRI scan in an inferior portion of the brain
is an increasingly important objective, as the use of DWMRI scans
increase. While the current use of isotropic EP DWMRI scans is likely
sufficient for superior locations in the cerebrum, the use of RD DWMRI for
inferior regions is worth considering.
References
1) Ross et al., Evaluation of Cancer Therapy Using Diffusion Magnetic Resonance
Imaging .Molecular Cancer Therapeutics 2003;2:581-7.
2) Theilmann et al., Changes in Water Mobility Measured by Diffusion MRI Predict
Response of Metastatic Breast Cancer to Chemotherapy. Neoplasia 2004;6:831-7.
3) Skare et al., Clinical Multishot DW-EPI Through Parallel Imaging With Considerations
of Susceptibility, Motion, and Noise. Magnetic Resonance in Medicine 57:881-890
(2007).
4) Sarlls et al., Isotropic Diffusion Weighting in Radial Fast Spin-Echo Magnetic
Resonance Imaging. Magnetic Resonance in Medicine 53:1347-1354 (2005).
5) Josien et al., Mutual-Information-Based Registration of Medical Images: A Survey,
IEEE Transactions on Medical Imaging, Vol. 22, No. 8, August 2003.
6) Simpson et al., Influence of Location and Extent of Surgical Resection on Survival of
Patients with Glioblasoma Multiforme: Results of Three Consecutive Radiation
Therapy Oncology Group (RTOG) Clinical Trials. Int. Journal. Radiat. Oncol. Biol.
Phys., Vol. 26 239-244 (1993).