Examination of Wavelet-Based Features for Congestive Heart Failure Classification Using SVM

One of the most common heart disease is congestive heart failure. In this study, the detail coefficients of RR interval data obtained from the discrete wavelet transform are applied for congestive heart failure (CHF) classification. The wavelet-based feature examined is a difference between the logarithms of variances of detail coefficients of RR interval data corresponding to two consecutive levels, referred to as ∆l. A feature vector used in CHF classification is formed by a pair of wavelet-based features, i.e., [∆m,∆n]. The classification is performed using SVM with a linear kernel function and its performance is validated using 10-fold cross-validation. From the computational results, the best performance on the CHF classification can be achieved using the feature vectors [∆2,∆3] and [∆1,∆2] where the best accuracy, the best sensitivity and the best specificity for the CHF classification are, respectively, 80.9324%, 81.4103% and 82.8938%.