Superior Vena Cava Computational Segmentation And Hypertensive Processes

A strategy for superior vena cava (SVC) three-dimensional segmentation is proposed using 20 cardiac imaging multi-layer computed tomography, for entire cardiac cycle of a subject. This strategy is global similarity enhancement technique based on and it comprises of pre-processing, segmentation and parameter tuning stages. The pre-processing stage is split into two phases called filtering and definition of a region of interest. These phases are preliminarily applied to end-diastole cardiac-phase and they address the noise, artifacts and low contrast images problems. During SVC segmentation, the region growing algorithm is applied to the pre-processed images and it is initialized using a voxel detected with least squares support vector machines. During the parameters tuning, the Dice score (Ds) is used to compare the SVC segmentations, obtained by the proposed strategy, and manually SVC segmentation, generated by a cardiologist. The combination of filtering techniques that generated the highest Ds considering the end-diastole phase is then applied to the others 19 3-D images, yielding more than 0.9 average Ds indicating an excellent correlation between the segmentations generated by an expert cardiologist and those produced by the strategy developed.