A new approach to detect congestive heart failure using detrended fluctuation analysis of electrocardiogram signals

The aim of this study is to evaluate how far the detrended fluctuation analysis (DFA) approach helps to characterize the short-term and intermediate-term fractal correlations in the raw electrocardiogram (ECG) signals and thereby discriminate between normal and congestive heart failure (CHF) subjects. The DFA-1 calculations were performed on normal and CHF short-term ECG segments, of the order of 20 seconds duration. Differences were found in shortterm and intermediate-term correlation properties and the corresponding scaling exponents of the two groups (normal and CHF). The statistical analyses show that short-term fractal scaling exponent alone is sufficient to distinguish between normal and CHF subjects. The receiver operating characteristic curve (ROC) analysis confirms the robustness of this new approach and exhibits an average accuracy that exceeds 98.2%, average sensitivity of about 98.4%, positive predictivity of 98.00%, and average specificity of 98.00%.