Compound extraction and fitting method for detecting cardiac ventricle in SPECT data

A new method for automatic extraction of the left ventricle of the heart from single photo emission computed tomography (SPECT) data is presented. The method involves extracting and fitting compound analytical (quadratic) surfaces which well-approximate the shape of the left ventricle. The approach exploits geometric shape constraints and utilizes Hough-based processing, hierarchical clustering, and least-squares fitting. Through extraction of the left ventricle, shape and size change analysis can be aided, which is useful for determination of cardiac health.