Automated Heart Wall Motion Abnormality Detection from Ultrasound Images Using Bayesian Networks

Coronary Heart Disease can be diagnosed by measuring and scoring regional motion of the heart wall in ultrasound images of the left ventricle (LV) of the heart. We describe a completely automated and robust technique that detects diseased hearts based on detection and automatic tracking of the endocardium and epicardium of the LV. The local wall regions and the entire heart are then classified as normal or abnormal based on the regional and global LV wall motion. In order to leverage structural information about the heart we applied Bayesian Networks to this problem, and learned the relations among the wall regions off of the data using a structure learning algorithm. We checked the validity of the obtained structure using anatomical knowledge of the heart and medical rules as described by doctors. The resultant Bayesian Network classifier depends only on a small subset of numerical features extracted from dual-contours tracked through time and selected using a filterbased approach. Our numerical results confirm that our system is robust and accurate on echocardiograms collected in routine clinical practice at one hospital; our system is built to be used in real-time.