Data processing for noninvasive continuous glucose monitoring with a multisensor device

Area of Science:

• Biomedical Engineering
• Medical Devices
• Biosensing Technology

Background:

• Impedance spectroscopy is a potential noninvasive method for continuous glucose monitoring.
• Skin and tissue impedance are influenced by factors beyond glucose levels, such as hydration.

Purpose of the Study:

• To investigate the effectiveness of impedance spectroscopy for blood glucose estimation.
• To evaluate the impact of personalized modeling on glucose prediction accuracy.

Main Methods:

• Principal component analysis (PCA) was used to analyze impedance spectra.
• Akaike's information criterion identified relevant variables.
• Linear least-squares modeling and Monte Carlo simulations were employed for model development and validation.

Main Results:

• PCA distinguished glucose-related impedance changes from hydration effects.
• A global model achieved R² = 0.60, while a personalized model reached R² = 0.71.
• Personalized models demonstrated a significant advantage over global models.

Conclusions:

• PCA effectively extracts glucose-related signals from skin impedance data.
• Linear models, particularly personalized ones, offer good predictive power for blood glucose.
• Personalization enhances the accuracy of impedance spectroscopy-based glucose monitoring.