Automated Pattern Classification for PCG Signal as a Novel Method for Clinical Decision Support System

Abstract

Auscultation pattern recognition also known as PCG Pattern classification was one of the efficient computer-based methods applied to medical decision making. PCG Pattern recognition generally is interpreted in two ways.  This work reports robust results with phonocardiogram PCG-signal pattern classification. Linear prediction analysis with basic agglomerative clustering techniques was applied to extract the spectral pattern from phonocardiogram signals, a relatively new technique. In this test, 35 PCG samples are classified correctly, except for seven samples; and 24 PCG samples correctly, except for three samples.  The efficiency of PCG spectral features classification has been confirmed experimentally to be integrated in automated auscultation computer aided diagnosis (Au CAD) systems. The more specific interpretation algorithms are limited to finding patterns in PCG signals or other related bio-signal activities. This work covers the new technique applied in basis of pattern classification for mitral regurgitation PCG signals to investigate different hemodynamics turbulences and stochastic blood flow patterns associated with cardiac circulation.

Keywords: Phonocardiography, Clinical Decision making, Pattern classification, Cardiovascular arrhythmia, Auscultation.