Inter-frame Echocardiography Feature Matching for Analyzing the Cardiac Motion

Automatic detection of systole and diastole ends is important clinical information that is difficult to obtain automatically from an echocardiography video. A diseased heart and a healthy heart have different dynamics. This can be tracked by observing the movements of cardiac muscles which surround all four cardiac chambers of the heart. In this paper, a feature based method is presented which automatically finds the pattern of the movement of heart and also point out the frames of systole and diastole. The cardiac muscle movements in echocardiography videos are tracked by five different feature extraction methods. After subjective evaluation, it was found that the KAZE features are most suitable compared to other feature extraction schemes. The technique of exhaustive searching is used to extract the relevant feature points from the next image frames. Outliers are detected After analyzing the spatial translation of those features in the subsequent image frame, various important information is extracted. The left ventricle region is particularly focused on obtaining various parametric information like systolic and diastolic ends, Systole duration and diastole duration, etc. Results indicate that the suggested method may estimate the said parameters rather closely to the clinical data.