Segmentation of Cardiac Chambers in 2D Echocardiographic Images by Using a New Atlas-Based Deformable Model

Abstract

Cardiovascular diseases (CVDs) are considered as the main reason of mortality around the world. Echocardiography is the most common imaging modality for diagnosis or treatment follow-up of CVDs. However, because of speckle-noise corruption and low resolution, segmentation of heart chambers in echo-images is a challenging endeavor. We previously proposed a probabilistic atlas as a prior model for the heart chambers in echo-images. In this paper, we propose a new active contour for the segmentation of cardiac chambers by using that digital atlas. Our deformable model effectively combines the global and local (patch-based) region-based energy functionals. Also, to extract all the cardiac chambers, we determine four active contours in every echo-image (one contour for each chamber in the four-chamber view). For simultaneously evolving all the curves, a coupling term is also added to the energy functional. Finally, the evolution equation of each active contour is computed through the Euler-Lagrange equation. Experimental results demonstrate that our method provides accurate solutions compared to expert delineations.