Title :
Shape analysis of brain ventricles using SPHARM
Author :
Gerig, G. ; Styner, M. ; Jones, D. ; Weinberger, D. ; Lieberman, J.
Author_Institution :
Dept. of Comput. Sci., North Carolina Univ., Chapel Hill, NC, USA
Abstract :
Enlarged ventricular size and/or asymmetry have been found markers for psychiatric illness, including schizophrenia. However, this morphometric feature is non-specific and occurs in many other brain diseases, and its variability in healthy controls is not sufficiently understood. We studied ventricular size and shape in 3D MRI (N=20) of monozygotic (N=5) and dizygotic (N=5) twin pairs. Left and right lateral, third and fourth ventricles were segmented from high-resolution T1w SPGR MRI using supervised classification and 3D connectivity. Surfaces of binary segmentations of left and right lateral ventricles were parametrized and described by a series expansion using spherical harmonics. Objects were aligned using the intrinsic coordinate system of the ellipsoid described by the first order expansion. The metric for pairwise shape similarity was the mean squared distance (MSD) between object surfaces. Without normalization for size, MZ twin pairs only showed a trend to have more similar lateral ventricles than DZ twins. After scaling by individual volumes, however, the pairwise shape difference between right lateral ventricles of MZ twins became very small with small group variance, differing significantly from DZ twin pairs. This finding suggests that there is new information in shape not represented by size, a property that might improve understanding of neurodevelopmental and neurodegenerative changes of brain objects and of heritability of size and shape of brain structures. The findings further suggest that alignment and normalization of objects are key issues in statistical shape analysis which need further exploration
Keywords :
Legendre polynomials; biomedical MRI; brain; harmonic analysis; image matching; image segmentation; medical image processing; surface fitting; 3D MRI; 3D connectivity; Legendre polynomials; binary segmentations; brain ventricles; dizygotic twin pairs; enlarged ventricular size; intrinsic coordinate system pairwise shape similarity; left lateral ventricles; mean squared distance; model-based segmentation; monozygotic twin pairs; morphometric feature; multiscale boundary description; neuroimaging; parameterization computation; polygonal surface mesh; psychiatric illness; right lateral ventricles; schizophrenia; series expansion; shape analysis; shape representation; single gradient-echo channel; spherical harmonics; supervised classification; ventricular shape; voxel-based objects; Approximation error; Biological system modeling; Brain modeling; Geometry; Humans; Shape measurement; Solid modeling; Statistical analysis; Topology; Visualization;
Conference_Titel :
Mathematical Methods in Biomedical Image Analysis, 2001. MMBIA 2001. IEEE Workshop on
Conference_Location :
Kauai, HI
Print_ISBN :
0-7695-1336-0
DOI :
10.1109/MMBIA.2001.991731