Volume Visualization of the Heart Using MRI 4D Cardiac Images

Abstract

This paper deals with a system for volume visualization of the heart using multiphase-multislice cardiac MRI data. The proposed system is based on a generalized 4D form of a fuzzy object extraction algorithm in order to distinguish voxels belonging to cardiac object from noisy points and surrounding tissues. The system is initialized by interactive selection of a pixel placed inside the cardiac muscle in a slice. In some cases, complementary information might be necessary for segmentation because of the similarity in grey level information between the heart and surrounding tissues. We applied an active contour model and a contour propagation technique to a variance image for a rough segmentation of epicardium. The fuzzy approach in combination with the use of a deformable model for isolation enables us to segment cardiac object without modifying voxel grey levels and preserving anatomical details. Fuzzy object extraction involves problems of enormous combinatorial complexity, but t his ca n be reduced by dynamic programming leading to practical algorithms for cardiac data sets. We have implemented these algorithms and tested their efficiency in preserving heart data during preprocessing. Different anatomical presentations of the heart have been used for this purpose, consisting of only a small number of slices per volume.