Automated detection of left ventricular boundaries for thallium-201 scintigrams

Computerized analysis of myocardial perfusion scintigrams requires identification of the left-ventricular (LV) boundaries. These boundaries are quite often manually outlined due to the high level of noise present in the images. The authors present an algorithm for extracting the LV boundaries from thallium-201 (Tl-201) scintigrams. The edge-extraction procedure is based on local-threshold selection by means of mixture-density identification. The method of moments is employed for identifying the mixture´s two normal components. Accordingly, the intensity level corresponding to the intersection point between the two normals is used to threshold the window under consideration locally. A comparison between the LV boundaries obtained by this algorithm and the manually outlined ones from the same Tl-201 scintigrams demonstrates that approximately 80% of the LV outline can be reliably generated