Human-in-the-loop AI predictive digital twin to extend virtual precision diabetes care between visits
https://www.nature.com/articles/s44401-026-00118-8
Publish Date: 2026-07-07 22:46:00
Source Domain: www.nature.com
-
Study Design: A randomized trial embedded within a parent RCT focusing on overweight/obese adults, examining the effects of an AI-enabled intervention during the second half of the study period.
-
Participants: Enrolled 60 adults with at least 18 years of age, including 40 individuals with type 2 diabetes (T2D). Participants were on either a ketogenic or low-fat, low-calorie diet.
-
Model Construction: Utilized a predictive digital twin involving an artificial neural network (ANN) trained with transfer learning to predict glucose and weight levels based on dietary intake, physical activity, and prior physiological measurements.
-
Human-in-the-loop Component: All AI-generated recommendations underwent review by a trained nurse or interventionist to ensure alignment with clinical guidelines and personalized for each patient.
-
Outcomes Focus: Measured changes in weight and glycemic stability during the intervention period. Weight change and glucose stability were assessed using ANOVA and Clark Error Grid analysis.
