3D heart segmentation and volumetry using deformable shape models

Abstract

Segmentation using deformable model-based approaches has become a common application in 3D medical image analysis, as it provides a fast and robust method to detect structures, e.g. organs, fully-automated after initialization. However, some earlier approaches in some cases did not converge in the segmentation process even on rather good datasets or showed major variations in the segmentation results. This work presents the application of a shape model enhanced with the possibility of free deformation with respect to local grey value distributions, used to automatically detect the end-diastolic and end-systolic left heart ventricle (LV) in 3D MRI images after a short user interaction, thus to calculate the heart stroke volume.