Identificação de pacientes com diabetes baseada na variabilidade da frequência cardíaca

Abstract

Diabetes mellitus (DM), usually referred to as diabetes, is a chronic disease characterized by hyperglycaemia and leads to specific long-term complications: retinopathy, neuropathy, nephropathy and cardiomyopathy. Analysis of heart rate variation (HRV), being a noninvasive tool, has become a popular method to assess the activitie of the autonomic nervous system (ANS). Heart rate (HR) are bio-signals that are in constantly changing. These changes may be an indication of current disease or serve as a pre-warning to imminent cardiovascular diseases. In this work, we analyse HRV signals from 360 normal and 360 diabetic subjects, using time domain, frequency domain and nonlinear techniques. Our results show that the indexes in the time domain (RRmean, SDNN, RMSSD, pNN50 and D index), in the frequency domain (VLF, LF, HF, HFnorm and LF/HF) and the nonlinear indexes (ApEn, SampEn, SD1, SD2, s, a1, FD, REC, DET, Lmean, Lmax and ShanEn) are clinically meaningful in the identification of patients with diabetes. The proposed diagnostic system classifies, DM patients and normal subjects, with an accuracy of 75:69%, specificity of 80:56% and sensitivity of 70:83%