Towards Reliable Early Diagnosis of Diabetic Cardiac Autonomic Neuropathy using Complexity features of ECG

Abstract

Cardiac Autonomic Neuropathy (CAN) is a serious and often underdiagnosed complication of diabetes mellitus, associated with a markedly increased risk of cardiovascular morbidity and mortality. The asymptomatic nature of early-stage CAN presents substantial challenges in clinical detection and timely intervention. This study introduces an integrated framework combining advanced signal processing and machine learning techniques for the early diagnosis of CAN using electrocardiogram (ECG) signals. Temporal dynamics of ECG segments were analyzed using Fractal Dimension and entropy-based measures to capture subtle perturbations in autonomic modulation. A hybrid diagnostic system encompassing both hardware acquisition and software analytical modules was developed, enabling robust, real-time assessment of cardiac autonomic function. Convolutional neural networks (CNNs) and fully connected neural networks (NNs) were employed for automated classification, achieving high diagnostic accuracy. The developed ECG-CAN device, incorporating these analytical techniques, has been provisionally patented, underscoring its innovation and translational potential. This approach demonstrates enhanced sensitivity and reliability for early CAN detection, paving the way for proactive risk management in diabetic populations.