Feature Extraction of the VSD Heart Disease based on Audicor Device Measurement

Abstract

This study intends to develop an automatic diagnosis system for Ventricular Septal Defect (VSD) based on Heart Sound (HS). The aim of this study is to create a Feature extraction of the VSD Heart Disease which will be use to classify between VSD HS and normal Heart Sound. The function of this system is preferred as a first-screening system before performing a more complex and costly medical test. The recording device for HS in this work is Audicor which has a feature for measuring Electro-Mechanical Activation Time (EMAT), Left Ventricular Systolic Time (LVST), third heart sound (S3) strength, capture heart sound, heart rate and Electrocardiograph (ECG). Since the function of Audicor is simply is to record the signals as mention above, to determine the VSD heart disease ultrasound is used, which is a medical standard instrument. Based on the Audicor measurement record, doctors analyzed the result and found that there were 29 cases of ventricular septal defect (VSD). Not only was the measurement for VSD patient, but also for a healthy person (normal heart sound). This system consists of several stages. They are as follow: taking heart sound data with Audicor and Doppler, segmentation, feature extraction by means of FFT and PCA. The data measurements were taking place at Chang Gung Memorial Hospital. This proposed system successfully created the feature of normal heart sound and VSD disease. This work will be continued with machine learning based and performed as a first screening tool to identify VSD Disease based on heart sound.