A Level-set Segmentation Approach for 4-D Cardiac Images

Cardiac diseases are one of the main causes of death in the World. This has motivated an important research effort aiming at the development of accurate tools for improving diagnosis and treatment. Recently, the Multi-Slice Computerized Tomography (MSCT) has emerged as a new source of 4D cardiac images that enables recording of cardiac shapes and their dynamical behavior during the cardiac cycle. This imaging technology requires the development of accurate techniques for analyzing and quantifying these images. This work presents the development of a software tool that enables a semi-automatic segmentation of cardiac cavities in MSCT images. The system core is a Level-Set algorithm. According to this algorithm, the contour is embedded as a zero level set of a higher dimensional level set function whose evolution described by a differential equation is performed considering features extracted from the images. The segmentation tool allows the user to define an initial rough segmentation by manual tracing of several contours in the 3D MSCT database. This approximate segmentation is improved using the level-set algorithm. The validation is performed by comparing the segmentation obtained using the level-set based algorithm with respect to the segmentation performed by medical experts. First results based on left ventricle extraction are promising.