Automatic View Recognition for Cardiac Ultrasound Images

Ultrasound images of the heart can be taken from many different angles. Diagnostic analysis of these images requires recognizing the pose of the heart so that important cardiac structures can be identified . We are concerned with the problem of automatically classifying cardiac ultr asound images with respect to what view of the heart they contain. Our solution to this problem has two novel aspects: first, a hierarchical classifier is con structed to classify an unknown view into corresponding windows, and then further classify it into one of the respective subclasses in the window; second, a simple dimension reduction approach is used to compress the information of many image features and to enhance the classification accuracy. Experiments on r ecognizing apical two-chamber, apical four-chamber, parasternal long axis and parasternal short axis views demonstrate the effectiveness of our approach.