Data Science Day @ Columbia University

In this demo we will showcase a novel way to utilize data collected through self-monitoring to predict health impact of future actions. Diabetes is a chronic illness affecting nearly 10% of Americans, and requires multiple adjustments to diet, exercise, and other lifestyle choices. Yet anticipating the impact of these adjustments can be challenging for both individuals with diabetes and their caregivers. We have developed Mealyzer, a mobile application that facilitates diabetes self-management by providing users with real-time, personalized, physiology-based forecasts of blood glucose levels based on prospective meals. Users record their blood glucose and meals by submitting images and descriptive text, and subsequently receive short-term blood glucose predictions that are generated in a data assimilation framework that uses a dual unscented Kalman filter to personalize mechanistic models of the glucose-insulin system. In this demo, we will showcase the Mealyzer mobile interface, and visualization of the forecasts generated with computational models.