Glucose
Xinyu Zhang / June 2025 (366 Words, 3 Minutes)
Unlocking Insights from CGM Data: Introducing Our New Interactive Browser
At our company, we are dedicated to revolutionizing metabolic health by combining continuous glucose monitoring (CGM) with a personalized AI coaching experience. A cornerstone of developing a truly intelligent coach is the ability to deeply understand and learn from large-scale, real-world health data.
To that end, we’re excited to showcase a powerful internal tool developed by our team: an Interactive CGM Browser. This tool was designed to bring a rich, multi-dimensional dataset to life—the “Continuous Glucose Monitoring Data from a Clinical Trial” available on PhysioNet.
About the Dataset
The PhysioNet CGM dataset is a treasure trove of information. It contains detailed records from a clinical trial, including not just CGM readings, but also:
- Biometric Data: Hemoglobin A1c, BMI, weight, and height.
- Lab Results: Fasting glucose, insulin levels, and more.
- Wearable Data: Heart rate, calories burned, and metabolic equivalents (METs) from activity trackers.
While incredibly valuable, the sheer volume and complexity of this data make it challenging to explore relationships and identify patterns using traditional spreadsheets or static plots.
Our Solution: An Interactive, Multi-Layered Browser
Our Interactive CGM Browser was built to solve this problem. It provides our data scientists, engineers, and clinical experts with a dynamic and intuitive way to visualize and analyze the entire dataset at once.
Key features of the browser include:
- An Integrated Overview: The main scatter plot allows us to instantly visualize relationships between key variables for all patients, such as the correlation between BMI and Hemoglobin A1c. Each point is clickable, serving as a gateway to deeper analysis.
- Detailed Patient Drill-Down: By clicking on any patient in the overview plot, the browser instantly loads their complete profile. This includes their demographic information, lab results, and a detailed time-series plot of their CGM, heart rate, and activity data.
- Flexible Time-Series Navigation: Users can effortlessly navigate through a patient’s entire monitoring period. The interface allows for zooming in on specific time ranges—from 6 hours to a full week—to examine granular events like post-meal glucose spikes or the impact of exercise.
- Dynamic Filtering: Our team can filter the entire dataset on the fly by patient ID, health status, or BMI category, making it easy to isolate and compare specific subgroups.
How This Tool Drives Our Mission
This Interactive CGM Browser is more than just a data visualization tool; it’s a critical asset that accelerates our research and development cycle. By enabling our team to rapidly explore, hypothesize, and validate ideas, it directly enhances our AI coach’s capabilities.
- For our Data Scientists: It helps uncover subtle correlations between lifestyle factors (like activity) and glucose responses, which are used to train more accurate predictive models.
- For our AI Team: It provides a rich source of real-world examples to build and refine the personalized feedback and recommendations our AI coach delivers.
- For our Clinical Experts: It offers a powerful way to review patient journeys and identify common patterns, ensuring our AI’s guidance is safe, effective, and clinically sound.
By building tools that allow us to deeply understand the nuances of metabolic health data, we are better equipped to empower our users with the personalized insights they need to take control of their health.