A wavelet-based soft decision technique for screening of patients with congestive heart failure

A wavelet-decomposition with soft decision algorithm is used to estimate an approximate power spectral density (PSD) of R–R intervals (RRI) of ECG data for the purpose of screening of congestive heart failure (CHF) from normal subjects. The ratio of the power in the low-frequency (LF) band to the power in the high-frequency (HF) band of the RRI signal is used as the classification factor. The trial data used for estimating of the classification factor consist of 15 CHF (patient) subjects and 12 normal sinus rhythm (NSR) or simply normal subjects. The performance of the algorithm is then evaluated on test data set, which consists of 17 CHF subjects and 53 NSR subjects. Both trial and test data are drawn from MIT database. The receiver operating characteristics (ROC) is used to determine the threshold value of the classification factor. Results are shown for different wavelets filters. The new technique shows a classification efficiency of 96.30% on trial data and 88.57% on test data. An FFT-based frequency domain screening technique is also implemented and included in this work for the purpose of comparison with the wavelet-based technique. The FFT-based technique shows an efficiency of classification of 99.63% on trial data and 81.42% on test data. The comparison is also done on short-term (5-min) recordings. The wavelet-based soft-decision technique shows also better results than the FFT-based technique.