Development of sensor system and data analytic framework for non-invasive blood glucose prediction

Area of Science:

• Biomedical Engineering
• Medical Devices
• Optical Sensing

Background:

• Current blood glucose monitoring relies on painful, invasive methods.
• There is a significant need for noninvasive, continuous glucose monitoring solutions.
• Existing noninvasive techniques face challenges with accuracy and external factors.

Purpose of the Study:

• To develop and validate a noninvasive glucose-monitoring device, niGLUC-2.0v.
• To assess the performance of finger and wrist sensor prototypes for glucose quantification.
• To overcome common ambiguities in noninvasive optical sensing for blood glucose.

Main Methods:

• Utilized near-infrared (NIR) spectroscopy with a 940 nm emitter and 900-1700 nm detector.
• Developed dual-sensor prototypes (finger and wrist) and collected data from 101 volunteers.
• Applied data pre-processing, exploratory data analysis (EDA), and ridge regression (RR) for analysis.

Main Results:

• niGLUC-2.0v achieved high accuracy: 99.02% (finger) and 99.96% (wrist).
• Low Mean Absolute Error (MAE) and Mean Square Error (MSE) were recorded for both prototypes.
• Bland-Altman analysis showed 98% of data within ±1.96 SD, and 100% within Clarke Error Grid Zone A.

Conclusions:

• niGLUC-2.0v demonstrates high accuracy and reliability for noninvasive glucose monitoring.
• The device effectively mitigates challenges like skin color and ambient light variations.
• niGLUC-2.0v is suitable for continuous, real-time blood glucose monitoring in medical and personal care settings.