Algorithm Produces Personalized Glucose Forecast for Patients with T2D
Machines were used to examine the glycemic impact of various meals to determine a glucose forecast.
HealthDay News — A mechanistic model based on Gaussian process models and data assimilation can produce a personalized, nutrition-based glucose forecast for individuals with type 2 diabetes, according to a study published online in PLOS Computational Biology.
David J. Albers, PhD, from Columbia University in New York City, and colleagues used 3 forecasting machines to examine the glycemic impact of different meals: data assimilation, which uses Bayesian modeling to infuse data with human knowledge in a mechanistic model to generate real-time, personalized forecasts; model averaging of data assimilation output; and dynamic Gaussian process model regression.
The researchers found that the data assimilation machine estimated states and parameters using a modified dual unscented Kalman filter, thereby personalizing the mechanistic models. To make a personalized model selection for the individual and their measurement characteristics, model selection was used. The data assimilation forecasts were assessed against actual postprandial glucose measurements that individuals with type 2 diabetes captured and against predictions that were generated by diabetes educators based on reviewing nutritional records and glucose measurements for the same individual.
"Our algorithm, integrated into an easy-to-use app, predicts the consequences of eating a specific meal before the food is eaten, allowing individuals to make better nutritional choices during mealtime," Albers said in a statement. "Now our task is to make the data assimilation tool powering the app even better."
Albers DJ, Levine M, Gluckman B, Ginsberg H, Hripcsak G, Mamykina L. Personalized glucose forecasting for type 2 diabetes using data assimilation [published online April 27, 2017]. Plos Comput Biol. doi:10.1371/journal.pcbi.1005232