Automated 3-D Tissue Segmentation Via Clustering

Abstract

Generation of 3-D tissue models from medical imagery is useful for surgical planning and computer simulations, but often requires some amount of manual effort. In this work, we use the clustering algorithm, DBSCAN, in concert with a 3-D buildup procedure to automatically generate 3-D surface models of the brain and lungs from computed tomography head and chest scans, respectively. Extensions to other tissue types such as heart and liver are demonstrated for contrast-enhanced imagery.