Detecting heart wall boundaries by tracking features in an echocardiogram sequence

An approach to extracting information from an echocardiogram sequence to the endocardial boundary as the heart undergoes deformation in a cardiac cycle is presented. Image velocities used as attributes of the edge segments are estimated by the image-flow approach, which relates the spatial intensity gradient to the temporal gradient. One component of the velocity can be directly computed, whereas the other must be recovered from other information. An algorithm based on the total-least-squares method is used to recover the lost components. Experimental results using an acquired echocardiographic image sequence are presented. Since the motion pattern in the image sequence is quite complicated, it is difficult to judge the results. The diastole effect is not apparent; however, the indicated velocities do not agree with the intensity variations