Automated Electrocardiogram Signals Based Risk Marker for Early Sudden Cardiac Death Prediction

Abstract

Sudden cardiac death (SCD) is one of the cardiovascular diseases that lead to millions of deaths worldwide every year. The aim of the present work is to propose a method for reducing the mortality rate of the SCD patients by an early prediction for SCD from the ECG signal. Normal and SCD MIT databases were used in this research work. One minute segments of ECG signals were segmented from MIT databases where these segments are ten minutes before sudden cardiac arrest (SCA) onset. The collected raw ECG signals were subjected to filter to remove the noise and then normalized. A frequency-domain feature and time-domain features were extracted from the Q-T segment, Q-T interval, R-R interval and QRS interval. The features were normalized to improve the performance of the classifier. Artificial intelligence classifiers; namely, K-nearest neighbor (KNN) and linear discriminant analysis (LDA) were used separately on SCD and normal ECG signals. The highest classification accuracy obtained for KNN and LDA are 97% and 95.5% respectively.