Assessment of Blood Glucose Measurement Using New Noninvasive Technology: Protocol and Methodology

Area of Science:

• Artificial Intelligence in Healthcare
• Biomedical Signal Processing
• Diabetes Mellitus Research

Background:

• Diabetes mellitus (DM) is a growing global health concern, with over 11.1% of adults affected, many undiagnosed.
• Early DM detection and management are crucial to prevent severe complications like vision loss, renal failure, and cardiovascular disease.
• Current invasive and costly screening methods limit widespread use, particularly in low-resource settings.

Purpose of the Study:

• To gather a diverse dataset for training and enhancing the Blood Glucose Evaluation and Monitoring (BGEM) machine learning algorithm.
• To improve the generalizability and robustness of the BGEM model for diabetes risk assessment and glucose monitoring.
• To evaluate BGEM performance across various demographic groups, including age, race, and skin types.

Main Methods:

• Recruited 885 adult participants (diabetic and non-diabetic) from the Greater Jakarta Area, Indonesia.
• Collected photoplethysmography (PPG) data using two wearable devices across four meal time points.
• Conducted laboratory blood glucose analysis (fasting and post-meal) and gathered anthropometric, physical activity, and medication data.

Main Results:

• Successfully enrolled 885 participants between June and October 2024.
• Acquired 8 photoplethysmography recordings per participant, alongside clinical measurements and questionnaires.
• Established a comprehensive dataset for evaluating BGEM's performance on diverse populations.

Conclusions:

• Outlined a methodology for assessing blood sugar profiles and evaluating the BGEM AI model using PPG data.
• The study focused on Indonesian participants, considering demographic variability and influencing factors for diabetes risk assessment.
• The generated dataset aims to validate the BGEM model's accuracy and reliability across diverse racial, risk factor, and skin-type groups.