Optimizing Multiscale Entropy Analysis for the Detection of Cardiac Pathology

The study investigates the use of the Rényi entropy algorithm with variable threshold and a signal temporal multiscaling approach for analyzing RR interval signals. The study involved 8-minute ECG recordings of 90 participants from the PhysioNet database and grouped into three groups: normal sinus rhythm, cardiac arrhythmia, and congestive heart failure. A time coarse-graining algorithm was used to obtain different temporal scales of the original signal. Rényi entropy probabilities of each scale-factored signal were calculated using a method of density based on sequences of the RR interval time series. ANOVA and post-hoc t-test were used to determine significant differences between the multiscaled Rényi entropy measures of the different groups of RR interval signals. The novel multiscaled Rényi entropy analysis provided enhanced significant discrimination between healthy and pathological (NSR vs. ARR and NSR vs. CHF) signals at post hoc t-test probability values of p < 5mathrm{x}10^{-6} and within pathological signals (ARR vs. CHF) at p < 5mathrm{x}10^{-} 3. The study concludes that applying a joint approach of cardiac signal temporal multiscaling and calculating its modified Rényi entropy with variable thresholding provides an optimized approach to identifying and separating healthy and pathological cardiac signals, and further supports the complex nature of the heart dynamics.