Cluster Analysis Framework for Novel Acoustic Catheter Stethoscope

Abstract

This study presents a cluster analysis framework for acoustic signals of the novel catheter stethoscope. The objective of the current study is to collect the blood flow sounds from a body site of a Yorkshire pig using the novel acoustic catheter stethoscope and further recognize any changes in the sinus rhythm patterns. Initially the collected blood flow sounds are preprocessed with noise cancellation and wavelet source separation to result in acoustic heart and respiratory pulses. Then, the extracted acoustic heart pulses are post processed to recognize the patterns of the sinus rhythm based on a novel feature extraction technique and cluster analysis. Finally, the developed framework is qualitatively and quantitatively validated by providing visual results of the clustering and by computing the clustering accuracy, sensitivity and specificity. The cross validation results show that the developed framework consistently recognizes the patterns of the sinus rhythms with an accuracy rate of 94.37%.