Robust MR Image Segmentation Using 3D Partitioned Active Shape Models

A 3D partitioned active shape model (PASM) is proposed in this paper to address the problems of 3D active shape models (ASM) brought by small training sets, which is usually the case in 3D applications. When numbers of training samples are limited, 3D ASMs tend to be restrictive, because the plausible area/allowable region spanned by relatively few eigenvectors cannot capture the full range of shape variability. 3D PASMs overcome this limitation by using a partitioned representation of ASM. Given a point distribution model, the mean mesh is partitioned into a group of small tiles. The statistical priors of tiles are estimated by applying principal component analysis to each tile, and the priors serve as constraints for corresponding tiles during deformations. To avoid the shape inconsistency introduced by the independent estimations between tiles, samples and deformed model points are projected as curves in one hyperspace, instead of point clouds in several hyperspaces. The deformed model points are then fitted into the allowable region of the model using a curve alignment scheme. The experiments on 3D human brain MRIs show that the 3D PASMs segment objects more accurately and are more robust to noise and low contrast in images than two other current active shape models. Furthermore, a study for the PASM's sensitivity to different initializations shows that PASMs perform stable when initialization positions change