SBIR Phase I: Automatic Classification of Magnetocardiograms

This Small Business Innovation Research (SBIR) Phase I project seeks development of novel machine learning capability for pattern recognition in magnetocardiography (MCG), which measures minute magnetic fields emitted by the electrophysiological activity of the heart. The company has developed a revolutionary measuring device for early identification/diagnosis of heart disease, inside regular, magnetically unshielded hospital rooms. However, interpretation of MCG recordings remains a challenge for cardiologists, since there are no databases from which precise rules could be deduced. Hence, there is an urgent need to automate and guide interpretation of MCG measurements, in order to minimize cardiologists' efforts to make meaningful diagnosis. The company will explore the application of automatic pattern recognition and classification schemes to MCG data. The goal is to develop a technique to accurately differentiate between abnormal and normal heart patterns and even to identify heart diseases. The differentiation will be performed using Support Vector Machines because they can handle high dimensional data and are especially suitable for a small number of samples. Since it is expected that different diseases will be located in separate clusters the investigative team plans to apply Self Organized Maps because they are suitable for multi-cluster formation. The proffered technology, which has the potential to deliver a diagnostic system for the detection of ischemia and coronary artery disease, could lead to the production of devices for detecting small disturbances in cardiac function and thus may warn doctors and patients of impending malfunction.