A Biomedical System Based on Artificial Neural Network and Principal Component Analysis for Diagnosis of the Heart Valve Diseases

• Published in: Journal of medical systems 2012-02, Vol.36 (1), p.61-72
• Main Author: Uuz, Harun
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Biomedical engineering; Cardiovascular disease; Child; Child, Preschool; Decision support systems; Decision Support Systems, Clinical - instrumentation; Decision Support Techniques; Female; Fourier Analysis; Health Informatics; Health Sciences; Heart Auscultation; Heart Valve Diseases - diagnosis; Humans; Male; Medical diagnosis; Medicine; Medicine & Public Health; Middle Aged; Neural networks; Neural Networks (Computer); Original Paper; Principal Component Analysis - methods; Principal components analysis; Signal Processing, Computer-Assisted - instrumentation; Statistics for Life Sciences; User-Computer Interface; Young Adult
• ISSN: 0148-5598; 1573-689X
• DOI: 10.1007/s10916-010-9446-7

Listening via stethoscope is a primary method, being used by physicians for distinguishing normally and abnormal cardiac systems. Listening to the voices, coming from the cardiac valves via stethoscope, upon the flow of the blood running in the heart, physicians examine whether there is any abnormality with regard to the heart. However, listening via stethoscope has got a number of limitations, for interpreting different heart sounds depends on hearing ability, experience, and respective skill of the physician. Such limitations may be reduced by developing biomedical based decision support systems. In this study, a biomedical-based decision support system was developed for the classification of heart sound signals, obtained from 120 subjects with normal, pulmonary and mitral stenosis heart valve diseases via stethoscope. Developed system was mainly comprised of three stages, namely as being feature extraction, dimension reduction, and classification. At feature extraction stage, applying Discrete Fourier Transform (DFT) and Burg autoregressive (AR) spectrum analysis method, features, representing heart sounds in frequency domain, were obtained. Obtained features were reduced in lower dimensions via Principal Component Analysis (PCA), being used as a dimension reduction technique. Heart sounds were classified by having the features applied as input to Artificial Neural Network (ANN). Classification results have shown that, dimension reduction, being conducted via PCA, has got positive effects on the classification of the heart sounds.