GlyTwin: Enhancing Digital Twin for Glucose Control in Type 1 Diabetes using Patient-Centric Counterfactual Treatments

Frequent and long-term exposure to hyperglycemia increases the risk of chronic complications, neuropathy, nephropathy, and cardiovascular disease. Existing continuous subcutaneous insulin infusion (CSII) and continuous glucose monitoring (CGM) technologies can only model specific aspects of glycemic regulation—such as predicting hypoglycemia and administering small insulin boluses. Similarly, current digital twin approaches in diabetes management are primarily focused on predicting glucose response to human behavior and insulin therapy. As a result, current technologies lack the ability to provide alternative treatment scenarios that could guide proactive behavioral interventions for optimal diabetes management. To address this gap, we propose GlyTwin, a novel computational framework that enhances capabilities of digital twin technologies by integrating counterfactual explanations to simulate optimal behavioral treatments for glucose control. GlyTwin generates counterfactual treatments by recommending adjustments to behavioral choices such as carbohydrate intake and insulin dosing to significantly reduce the occurrences and duration of hyperglycemic events. Additionally, GlyTwin incorporates stakeholders’ preferences into its intervention-generation process and ensures that the tool itself is personalized and patient-centric. We evaluate GlyTwin on AZT1D, a new dataset that we have constructed by collecting longitudinal data from 21 patients with type 1 diabetes (T1D) on automated insulin delivery (AID) systems, each monitored for 26 days. Results show that GlyTwin outperforms state-of-the-art methods for generating counterfactual explanations with 76.6% valid explanations and 86% effectiveness in preventing hyperglycemia when compared against historical data.

Preprint

ArXiv