A Cost-Effective IoT-Based Non-Invasive Glucose Monitoring System with Individualized Calibration for Scalable Biomedical Device Production

Abstract

This study presents a cost-effective, IoT-enabled, non-invasive glucose monitoring system that employs photoplethysmography (PPG) biosignals and individualized calibration to enhance measurement accuracy and long-term stability. The system integrates the MAX30101 Pulse Oximeter and Heart Rate Sensor for high-fidelity biosignal acquisition and the MAX32664 Biometric Sensor Hub for noise reduction and signal processing. An Arduino Uno R4 WiFi microcontroller with cloud-based storage and real-time dashboards handles data. A personalized polynomial regression model was developed for a single user, as a proof-of-concept. The model uses heart rate (HR) and peripheral capillary oxygen saturation (SpO2) as input features to predict glucose levels, while accounting for physiological variability. Model evaluation demonstrated strong predictive performance, achieving an R² of 0.8576 during training and 0.789 during testing, with a mean absolute error (MAE) of 6.58 mg/dL, a root mean square error (RMSE) of 7.75 mg/dL, and a mean absolute relative difference (MARD) of 7.13%. The achieved accuracy is comparable to values reported in the PPG-based non-invasive glucose monitoring literature, while the use of individualized calibration in this study provides a key advantage: tailoring the regression model to each user’s specific physiological characteristics, thereby minimizing the need for frequent recalibration and enhancing long-term stability. To the best of the authors’ knowledge, this is the first low-cost IoT-based glucose monitoring system that combines PPG sensing with individualized calibration. This low-cost proof-of-concept can be generalized and scaled through a structured pathway that progresses from multi-subject clinical validation to engineering design transfer with factory-level verification, followed by broader deployment supported by regulated quality and approval processes.