FRACTAL DIMENSION TECHNIQUES FOR ANALYSIS OF CARDIAC AUTONOMIC NEUROPATHY (CAN)

Abstract

Identifying Cardiac Autonomic Neuropathy (CAN) in the early stages of proliferation demands more prominent techniques with a reliable significance of identification. CAN being a subclinical consequence that is the leading cause of death in individuals with diabetes mellitus (DM), which is common among one in four people above an average age of 45 years, calls for a more dependable technique for analysis. This study investigates the complexity in prominent time segments (RR, QT and ST) of ECG using different entropy measures and four nonlinear fractal dimension (FD) measures including box counting, Petrosian, Higuchi’s and Katz’s methods. Measures of statistical significance were implemented using Wilcoxon, Mann–Whitney and Kruskal–Wallis tests. The results of the study provide an original approach to diagnostics that reveals the fact that, instead of analyzing the signal running for the whole length, complexity measures can be achieved, if the intervals of the signal are studied including a combination of features rather than any one feature considered for diagnosis. A significance level of [Formula: see text] is achieved in more segments of ECG considered at intervals of time compared to one data recorded at the 20th minute between CAN+ and CAN− groups for both FD and entropy. Neural Network (NN) classification shows the accuracies of 84.61% and 60% in FD and entropy, respectively, computed every fifth minute. The accuracies from the model for the data collected at the 20th minute for FD and entropy are 50.22% and 30.33%, respectively, between the groups.