Computerized heart valve diagnosis using a digitized heart sound.

摘要

Over the past two decades stethoscope auscultatory skills of primary care physicians have declined. One study showed that general practitioners correctly diagnose heart sounds only 30% of the time (Ortiz-Neu C., Tenenbaum J., 2001). The ultimate goal of this research was to produce a software program that will aid the medical professional in cardiac diagnosis. The resultant algorithm analyzes the heart sound and presents whether or not the patient should be referred to a cardiology specialist. A similar scheme is used in automated EKG readings where readings are analyzed and a few alternative diagnoses are suggested. The idea is to help general practitioners, internists, physician's assistants and nurses with the initial diagnosis by providing more information to physicians who are stage heart disease, which will reduce unnecessary diagnostic testing and save money.; Auscultation-based heart sounds from cardiac patients were collected using a digital stethoscope. Diagnoses were determined from the current gold standard, the Echocardiogram. Data preprocessing, including segmenting the heart sounds into heart cycles composed of S1, systole, S2 and diastole and resampling the data to a standard heart rate of 70 beats per minute were engineered. Multi-dimensional decision spaces were determined using a Principal Components Analysis and the Singular Value Decomposition.