A significant percentage of young children present cardiac murmurs. However, only one percent of them are caused by a congenital heart defect; others are physiological. Auscultation of the heart is still the primary diagnostic tool for judging the type of cardiac murmur. An automated system for an initial recording and analysis of the cardiac sounds could enable the primary care physicians to make the initial diagnosis and thus decrease the workload of the specialised health care system. The first step in any automated murmur classifier is the identification of different components of cardiac cycle and separation of the murmurs. Here we propose a new methodological framework to address this issue from a machine learning perspective, combining Independent Component Analysis and Denoising Source Separation. We show that such a method is rather efficient in the separation of cardiac murmurs. The framework is equally capable of separating heart sounds S1 and S2 and artifacts such as voices recorded during the measurements.
|Title of host publication||Independent Component Analysis and Blind Signal Separation. Proceedings (Lecture Notes in Computer Science)|
|Publication status||Published - 2006|
|Event||6th International Conference, ICA 2006 - Charleston, SC, United States|
Duration: 2006 Mar 5 → 2006 Mar 8
|Conference||6th International Conference, ICA 2006|
|Period||2006/03/05 → 2006/03/08|