Statistical 3D shape-model guided segmentation of cardiac images
Y. Shang, Olaf Dössel · 2005
A novel scheme for the segmentation of 4-D MR cardiac images has been introduced in this paper. 3-D spatially hierarchical expressions of the statistical shape models for the cardiac chambers are constructed through principal component analysis (PCA) of the manually segmented training set. The difficult problem of landmarking before PCA is effectively solved through gradient vector flow guided 3D active surface method. The hierarchical shape models are then used as constraint forces in the segmentation process. The system has been tested on 3D MR cardiac images. The segmentation results of the presented system are better compared with the active surface method, and are comparative to the manual segmentation results.