Non-Invasive Blood Glucose Estimation Based on Machine Learning Algorithms Using PPG Signals

Managing diabetes effectively requires accurate monitoring of blood glucose levels. Traditional invasive methods for such monitoring can be cumbersome and uncomfortable for patients. This study introduces a non-invasive approach to estimate blood glucose levels using photoplethysmography (PPG) data. It focuses on fasting blood glucose prediction using wrist PPG signals and explores various PPG-based features, including AC to DC ratio component of PPG signal and the ratio of different wavelength AC/DCs. The study highlights feature selection to improve model accuracy and efficiency by eliminating redundant features and addressing the challenges required to accurately capture glucose trends with PPG signals. Machine learning algorithms, including random forest, CatBoost, XGBoost, and LightGBM, were employed to analyze PPG signals and estimate the corresponding glucose levels. This non-invasive, continuous monitoring approach can significantly enhance diabetes management by reducing the need for frequent blood sampling, improving patient compliance, and providing real-time blood glucose level insights.