Evaluation Of Techniques For The Detection Of The Left Ventricular Cavity In X-Ray Images Of Cardiac Angiography

In this work, the performance of three (3) techniques designed for left ventricular cavity detection applied in cardiac angiography images are compared. The techniques obtain the necessary information for quantifying ventricular function, based on a) linear approximation, b) deformable models (snake), and c) clustering techniques and functional relations. Comparison is established following a methodology that allows quantifying the final shape obtained by each proposed method with respect to the real shape traced by a cardiologist. Experimental results show that techniques based on deformable models are more robust than the other two proposed techniques. These models accurately match the smoothness and curvature features of the left ventricular shapes. However, the snake and the linear programming technique require an initial set of points traced manually near the shape to be extracted, while the clustering techniques and the functional relations approach are able to extract automatically the ventricular shape.