An Automatic Random Walk Based Method for 3D Segmentation of The Heart In Cardiac Computed Tomography Images

Cardiac Computed Tomography (CT) is an important imaging modality for diagnosing cardiovascular disease. Heart segmentation from CT scans is a prerequisite for clinical interpretation, providing critical information for quantitative analysis of the heart and coronary structures. This paper presents a fully automated algorithm for heart segmentation based on the random walk model and automated initial seeds detection. The performance of the method is evaluated on a cardiac CT dataset of 58 subjects using manual segmentation as a reference standard. We show the results of our method are similar to the reference standard and are comparable to the results from a commercial product. We also demonstrate our automated method can reliably segment other tissue structures in cardiac CT images.