In-human testing of a non-invasive continuous low-energy microwave glucose sensor with advanced machine learning...

Area of Science:

• Biomedical Engineering
• Microwave Sensing Technology
• Wearable Health Devices

Background:

• Continuous glucose monitoring (CGM) is crucial for diabetes management, but current methods often require invasive finger pricking.
• Developing noninvasive sensing technologies is essential to improve patient comfort and adherence.
• Microwave sensing presents a promising avenue for biomedical analyte detection.

Purpose of the Study:

• To introduce a compact microwave planar resonator-based sensor for noninvasive glucose monitoring.
• To validate the sensor's performance through in vivo, in vitro, and clinical trial testing.
• To enhance the system with machine learning for predictive glucose analysis and artifact detection.

Main Methods:

• Design and implementation of a compact planar resonator microwave sensor.
• Conducting in vivo and in vitro tests using a microfluidic channel system.
• Performing clinical trials to assess real-world performance.
• Integrating machine learning algorithms for data analysis and prediction.

Main Results:

• The sensor demonstrated reliable operation in various testing settings.
• A high linear correlation (R² ≈ 0.913) was achieved between sensor response and blood glucose levels.
• The system provided real-time glucose level readings.
• Machine learning enabled accurate prediction of glucose level variations and artifact identification.

Conclusions:

• The proposed microwave planar sensor offers a viable noninvasive solution for continuous glucose monitoring.
• The integration of machine learning enhances predictive capabilities and data reliability.
• This technology paves the way for customized, learning-enabled wearable glucose monitoring devices.