A Spectral Exponent-Based Feature of RR Interval Data for Congestive Heart Failure Discrimination Using a Wavelet-Based Approach

Abstract

Congestive heart failure (CHF) is a common chronic condition which affects several millions of people around the world. Heart rate data which are crucial information needed for diagnosis and classification of CHF are similar to several physiological signals that exhibit an extraordinary range of patterns and behaviors. In this study, the spectral exponents of RR interval data of two groups of subjects, i.e., subjects with CHF and subjects with normal sinus rhythm, obtained using a wavelet-based approach are examined where the second order Daubechies wavelets are used. The spectral exponent of RR interval data is determined from a slope of logarithm of variance of wavelet coefficients (log2var(d m,n )) versus levels of wavelet-based decomposition (m) graph. In particular, the spectral exponent is estimated from the levels of wavelet-based decomposition ranges between m = 1 and m = 3 corresponding to finer-scale components of RR interval data. The minimum of spectral exponents of epochs of RR interval data is proposed as a quantitative feature for discriminating a subject with CHF. The computational results show that the subjects with CHF can be perfectly discriminated from the subjects with normal sinus rhythm using the spectral exponent-based feature. Furthermore, the perfect CHF discrimination can be achieved using the RR interval data with epoch size as short as 128 beats (approximately 2 min).