EPILEPSY RESEARCH GROUP
Institute of Neurology, UCL
MRreg Image Registration and Processing Package
Contents:
1. General Description
MRreg is a complete software environment for
the registration and analysis of brain MRI data. It was developed primarily
for the automatic quantification of changes in serial MRI volume scans
of the brain, in particular to quantify signal changes in low-grade
brain lesions (see illustration below). However,
MRreg is also useful for the following purposes (among others):
Detect and quantify volume changes;
Measure volumes (whole brain or sub-structures)*;
Register scans from different subjects in a common space.
Register volume scans from the same subject but with different contrasts.
The precision and accuracy of the registration in
MRreg have been fully validated, with a registration
accuracy for test objects of 0.02mm and a registration 'consistency' of
0.06mm in living subjects (please consult the 1998 papers in Medical
Physics and Medical Image Analysis in the publications
list below).
MRreg consists of the main following components:
1. Fully automatic determination of the rigid body transformation
matrix (with or w/o linear scaling) that best relates two data sets;
2. Sinc-based or linear interpolation of the matched
scan;
3. Intensity matching and subtraction;
4. Non uniformity correction of the difference volume;
Specific analysis sub modules:
5. Automatic random (Gaussian) noise quantification in
difference images;
6. Automatic quantification of subtle signal changes
in serial scans;
7. Volume measurements by region growing or manual delineation*.
MRreg offers most standard image display
features, eg intensity windowing, zooming, etc, plus:
Side-by-side display of target, match and matched data
sets;
Interactive perpendicular reformating (eg: sagittal or
axial slices from coronal);
Voxel intensity value display;
Linked cusrors.
MRreg is accompanied by an automatic brain
extraction tool, called Exbrain.
*MRreg has facilities for saving
and retrieving manually drawn traces, region growing seeds and threshold
levels, and for the automatic calculation of areas and volumes.
2. Example
The following illustrates the quantification of signal changes in a lesion
in a series of five volume scans.
The top left image is a slice from the baseline volume scan. The
second column contains slices from from subsequent scans after registration.
The third column contains slices from the difference volume, after intensity
matching. The fourth column contains the segmented differerences, with
red representing significant signal decrease and green, signal increase.
The latter is obtained by automatic thresholding of the structured part
of the difference image and comparison with an anatomical map of changes
and artefacts in normal subjects (derived from serial scan pairs of 20
normal subjects).
3. Main Publications
(as of 29/01/99)
Lemieux L, Wieshmann UC, Moran NF, Fish DR, Shorvon SD,
"The Detection and Significance of Subtle Changes in Mixed-signal Brain
Lesions by Serial Scan Matching and Spatial Normalisation," Medical Image
Analysis 2(3): 227-242 (1998). Web version (Oxford University Press)
Lemieux L and Barker GJ, "Measurement of Small Inter-scan
Fluctuations in Voxel Dimensions in Magnetic Resonance Images Using Registration,"
Medical Physics 25(6): 1049-1054 (1998).
Moran NF, Lemieux L, Maudgil DD, Kitchen ND, Fish DR, Shorvon SD, "Analysis of temporal lobe resections in MR images,"
Epilepsia 40(8): 1077-1084 (1999).
Tebartz van Elst L, Woermann FG, Lemieux L, Trimble, MR,
"Association between Left Amygdala Volume and Depression Score in Patients
with Temporal Lobe Epilepsy - a MRI Volumetric Study," Biological Psychiatry,
accepted for publication.
Tebartz van Elst L, Woermann FG, Lemieux L, Trimble, MR,
"Affective aggression in patients with temporal lobe epilepsy - a quantitative magnetic resonance imaging study of the amygdala," Brain,
accepted for publication.
Wieshmann UC, Woermann FG, Lemieux L, Free SL, Bartlett
PA, Smith SJM, Duncan JS, Stevens JM and Shorvon SD, "The Development of
Hippocampal Sclerosis: a Serial MRI Study over a 58 Month Period in a Patient
After Status," Epilepsia 38: 1238-1241 (1997).
Lemieux L, "Segmentation and Estimation of Gaussian Noise
and Structure in Difference Images of Co-registered MR Imaging Scan Pairs,"
Radiology, 205(P): 167 (1997).
Lemieux L, "Automatic Estimation of the Random Noise Level
in Difference Images from Matched Brain Scans Based on the Segmentation
of Structure," Proc. ISMRM 1998: 2082.
4. Contact
Louis Lemieux,
PhD
Lecturer in Physics
Epilepsy Research Group
Institute of Neurology, UCL
33 Queen Square,
London WC1N 3BG
United Kingdom