Noninvasive Reflective Optical Blood Glucose Measurement System Based on Red/Green and Infrared/Green Light Amplitude Ratios

This study innovatively utilizes a multi-wavelength photoplethysmography (PPG) approach to improve the non-invasive measurement of blood glucose levels by calculating amplitude ratios (R/G and IR/G) using green light as a baseline for calibrating blood flow measurements in the test area. By integrating amplitudes of PPG features with PPG amplitude ratios, we achieved a root-mean-square-error (RMSE) of 7.18±5.66 mg/dL for blood glucose estimation in 92 participants (76 diabetic patients and 16 healthy controls). The results demonstrated that even with simple linear regression methods, the model attained over 90% accuracy in Clarke plot analysis, highlighting the potential clinical applicability of this approach. For feature point detection, we proposed using curvature detection to identify the dicrotic notch and diastolic peak features, which tend to be less pronounced in diabetic patients. The addition of curvature detection specifically addresses the challenge of feature recognition in patients with vascular complications, a common issue in diabetes. The straightforward nature of this approach, coupled with the robustness of traditional regression analysis, suggests that further refinements using advanced AI techniques could significantly improve the predictive accuracy, paving the way for more precise and reliable diabetic monitoring technologies.