Oxidative Stress Markers Identify Cardiac Autonomic Neuropathy Progression: Applying Machine Learning Methods

Abstract

This study aims to highlight the association between oxidative stress and cardiac autonomic neuropathy (CAN) using machine learning algorithms for risk prediction. Oxidative stress is a significant factor in chronic diseases. Data from 2,621 participants were provided by the DiabHealth diabetes complications screening clinic at Charles Sturt University (CSU) for analysis, spanning the years 2002 to 2015. The oxidative stress markers considered in this study were 8-isoprostane, 8-hydroxy-2‘-deoxyguanosine (8-OHdG), reduced glutathione (GSH), oxidized glutathione (GSSG) and glutathione redox ratio (GSH/GSSG). Machine learning methods, including Random Forest and Logistic Regression, were employed to develop two multi-class and one binary model. For ROC-AUC, all models achieved relatively high values where “Definite” in model 1 is 0.82, “Normal” in model 2 is 0.81, and “Abnormal” in model 3 is 0.81. The findings underline the potential of integrating machine learning methods in CAN prediction, offering substantial improvements over traditional methods. By exploring novel multi-class models and unveiling the capabilities of the random forest classifier, this research establishes a robust foundation for future investigations.