Mapping data to virtual patients in type 1 diabetes

In this work, a new approach to represent dynamic variations in type 1 diabetes (T1D) leveraging field-collected data (glucose monitor, insulin pump and meal records) is presented. The proposed methodology consists in identifying a time-invariant model structure of the glucose metabolism along with a variability component (VC) that captures daily changes in insulin sensitivity (IS). Performance is tested using both synthetic and real data, evidencing the ability of the methodology to replicate and predict outcomes under diverse conditions. This reconstruction-replay framework lays the foundation for a viable replay engine that can represent a powerful tool for precision medicine in T1D.