Classification of arteriovenous fistula stenosis using shunt murmurs analysis and support vector machine

Abstract

Hemodialysis patients are generally provided with a shunt, but problems such as stenosis of the blood vessel can occur. It is effective for hemodialysis patients to check their own shunt function by listening to shunt murmurs. However, manually judging shunt function is difficult and requires experience. Therefore, automatic classification of shunt functions to analyze shunt murmurs could be an effective method for checking shunt function. In this study, we propose a method to classify shunt stenoses using support vector machine (SVM). We use the resistance index (RI) obtained from the ultrasonic diagnostic equipment as a class label and the normalized cross correlation coefficient, the ratio of the frequency power and Mel-Frequency Cepstral Coefficients (MFCC) as the feature learned by the SVM classifier. As a result, the accuracy of classification of RI by SVM was lower than that obtained by human judgment.