ABSTRACT
Identification of myocardial muscle dysfunction by quantitative analysis allows a reliable diagnosis of cardiovascular diseases. Machine learning techniques have also been used for cardiac cavities segmentation such as Zheng et al. where a two stages method is proposed. The first stage performs anatomical structure localization, and in the second stage the boundary delineation is attained. The segmentation of cardiac images is a difficult task because of several problems such as: complexity and variability of the heart movement, the low contrast exhibiting by the objects in these images, contamination with noise and presence of artifacts that may arise during the acquisition and/or reconstruction of the images. Results obtained by the level set algorithms for the segmentation of the left ventricle in 4-D Multi-Slice CT images, are dependent on the initialization. The Caselles algorithm provides high Dice similarity coefficients and their implementation using sparse field techniques provides low computation times.
