Effect of biosignal preprocessing and recording length on clinical decision making for cardiac autonomic neuropathy

Abstract

Early identification of cardiac autonomic neuropathy (CAN) leads to better treatment outcomes. Heart rate variability (HRV) analysis allows identification of CAN but is sensitive to the length of recording and the presence of artifacts and ectopics, requiring preprocessing and consideration of length of recording. RR intervals from 10-second and 5-minute ECG recordings from patients with no CAN, early CAN and definite CAN were preprocessed using adaptive filtering with the controlling parameter c set at 0.2, 0.5 and 0.8, and time and frequency domain HRV analysis applied. Early CAN and definite CAN required different setting of c with respect to the length of recording. The 5-minute recording with c=0.2 provided the best results using RMSSD for normal versus eCAN (p=0.0007) and for eCAN versus dCAN (0.019). Clinically, this has potential use in general practice for screening patients at risk, newly diagnosed with diabetes, or for follow-up during the course of diabetes.