Diagnosis of Congestive Heart Failure from HRV signal using SVM classifier and Patient Specific cross validation

Congestive heart failure (CHF) is popularly known fatal cardiac disease that occurs when pumping action of heart is lower than normal case. The purpose of this study is to the accurate diagnosis of CHF by improving classifier performance with effective features extraction and cross validation approach. The identification of significant features in electrocardiogram is highly important to detect congestive heart failure. Therefore ,this paper introduces a classifier based automated detection scheme with a novel approach of feature extraction from Heart rate Variability (HRV) signal for early prediction of CHF. The dynamical characteristics of HRV signal is analysed by computation of Largest Lyapunov Exponent. The statistical features are also evaluated to capture crucial variation in HRV signal to distinguish the abnormal and normal heart condition. The extracted features are subjected to Support Vector machine (SVM) classifier for automated discrimination of CHF from normal ECG signal. Experimental results evaluate the performance of extracted features and estimate the accuracy of the classification using the 10 fold cross validation and patient specific cross validation approach. Our experiment is validated by ECG data of normal and CHF subjects from Physionet database. The proposed system is efficient to the detect CHF with an average accuracy of 98.75%,, sensitivity 98.38%, values and 98.94%. Based on comparative study with the existing scientific research work to diagnose CHF, our proposed approach is found to be reliable and efficient for CHF diagnosis