Student Project

Harnessing Multimodal Biosensors and Novel Metrics for Predictive Diabetes Management

Team members
Joseph Accurso

This INFO 290T project investigates the use of wearable biosensors and machine learning to enhance predictive management of Type 1 diabetes. By collecting continuous biometric data—including pulse rate, pulse rate variability, body temperature, respiratory rate, and electrodermal activity—from an Empatica EmbracePlus, and aligning it with real-time glucose measurements from a Dexcom G7 continuous glucose monitor, the study explores correlations between physiological signals and blood glucose trends.
Advanced modeling techniques revealed a particularly strong predictive relationship between pulse rate variability and glucose levels, suggesting potential for more autonomous, non-invasive glucose monitoring systems with predictive abilities. Additionally, the project introduces the Average Rate of Change (AROC) as a novel metric to complement traditional A1c measurements, offering a dynamic perspective on glycemic variability. These findings advocate for integrating multimodal biosignals into diabetes care to improve real-time insights and support more personalized treatment strategies.

Last updated: May 13, 2025