Quantifying regional left ventricular function using spatio-temporal tracking techniques
Increasingly, diagnosis of cardiac disease, relies on computer processing of images to aid decision making. In this thesis, we use echocardiography, which is the most widely used cardiac imaging modality to study the motion of the left ventricle. Currently, clinical reporting of echocardiography examinations is operator-dependent and largely qualitative. Commercially available software does not track the left ven- tricle. Also, it does not provide quantification of regional function. This thesis establishes a framework for the quantitative regional analysis of left ven- tricular function. The endocardial and epicardial contours are automatically tracked during the cardiac cycle. A quantitative measure of regional endocardial wall excur- sion and myocardial thickening, based on a 16-segment model of the heart, is then obtained based on these boundaries. The new tracking framework is based on Kalman filtering which makes a single pre- diction as to the position of the boundary on the next frame. We develop a mea- surement model for the endocardial border, the tissue/blood interface, and the epi- cardium, the tissue/tissue interface. Having tracked the endocardial and epicardial boundaries, we introduce an interpretational space which provides clinically mean- ingful regional quantitative measures of left ventricular function. We illustrate all the concepts on one example. We apply the ideas developed to stress echocardiography, in a small retrospective clinical test.