Noninvasive Blood Glucose Monitoring with Machine Learning Enhanced Transmittance Spectroscopy.

Area of Science:

• Biomedical Engineering
• Spectroscopy
• Machine Learning in Healthcare

Background:

• Diabetes mellitus is a global health concern requiring frequent blood glucose monitoring.
• Current finger-prick methods for blood glucose level (BGL) monitoring are invasive and inconvenient.
• Noninvasive glucose sensing technologies are highly sought after for improved patient compliance and continuous monitoring.

Purpose of the Study:

• To develop and validate a noninvasive system for estimating blood glucose levels (BGL) using transmittance spectroscopy and machine learning.
• To identify the optimal wavelength for noninvasive glucose measurement with minimal interference.
• To assess the accuracy and reliability of the developed system for diabetes management.

Main Methods:

• Multispectral transmittance measurements were taken at 650, 808, and 940 nm.
• Light absorption simulations and in vitro experiments identified 940 nm as the optimal wavelength.
• An in vivo system utilized transmittance spectroscopy and an XGBoost machine learning model trained on 200 clinical samples.

Main Results:

• The XGBoost model achieved a high accuracy with an R-squared score of 0.94 and RMSE of 23.92 mg/dL.
• Clarke Error Grid Analysis (CEGA) indicated 92.5% of predictions in Zone A and 7.5% in Zone B.
• No critical errors were observed, demonstrating the system's robustness and safety.

Conclusions:

• The proposed noninvasive transmittance spectroscopy method shows significant potential for accurate glucose monitoring.
• This technology offers a practical, low-cost, and reliable solution for continuous glucose monitoring in diabetes management.
• Further development could lead to improved noninvasive diabetes care and patient outcomes.