Blood Glucose Research
From Your Wrist
Help us pioneer non-invasive blood glucose monitoring using Samsung Watch 7/8 and PPG sensors. Be part of cutting-edge research at the intersection of wearables and health AI.
What We're Building
We're developing an AI model to estimate blood glucose levels from consumer wearables, specifically the Samsung Galaxy Watch 7 and 8. Using advanced PPG (photoplethysmography) signal processing and machine learning, we aim to provide personalized glucose insights without invasive sensors.
This research builds on our work at Track Our Hearts, where we've successfully implemented real-time ECG analysis, HRV monitoring, and multi-device sensor fusion.
What We Need From You
Samsung Watch 7 or 8
Own and be willing to use our research app on your device
Freestyle Libre Data
Share 2+ weeks of continuous glucose monitoring data
Fingerstick Measurements
Days 1-4: 6-8 measurements per day
Days 5-11: 4 measurements per day
Record exact values and timestamps each time
How It Works
Data Collection (2 weeks)
- • Wear your Samsung Watch continuously
- • Use our app to sync PPG data
- • Record fingerstick measurements
- • Share Freestyle Libre data
- • Data collection only (no estimates yet)
Model Training & Personalization (2-4 weeks)
Starts approximately 2 weeks after Phase 1 begins
- • We train AI models on your data
- • Personalized glucose estimation
- • Continuous model improvement
- • Real-time insights in the app
- • Help validate the technology
⚠️ NOT FOR MEDICAL USE - Research and Informational Only
What You Get
Personalized glucose insights trained specifically on your physiology
Free early test access to Track Our Hearts when it opens
Be a pioneer in non-invasive continuous glucose monitoring
Contribute to research that could help millions manage their health
Important: Research Only
This is experimental research. The glucose estimates provided are NOT intended for medical diagnosis or treatment decisions.
Do not use this technology to adjust insulin doses, medication, or make any medical decisions. Always consult with your healthcare provider and use FDA-approved devices for medical purposes.
By participating, you acknowledge this is a research study with no guarantee of accuracy or clinical validity.
The Technology
Our approach uses advanced signal processing on PPG (photoplethysmography) data from your watch's optical sensors. We extract 37 features including:
- • Heart rate & HRV
- • Pulse wave analysis
- • Perfusion metrics
- • Frequency domain features
- • Coherence measures
- • Entropy analysis
- • Multi-task learning
- • Subject embeddings
- • Uncertainty quantification
Create an account or sign in to begin