Differentiable T1D Simulators

This repository contains various simulators/dynamical systems for modeling Type 1 Diabetes (T1D) dynamics. We implement our models in Jax and PyTorch, allowing for easy differentiation of the dynamics.

Examples

See the examples folder for examples of how to generate data using our simulators.

Installation

1. Clone the repository: git clone git@github.com:KeAWang/differentiable_t1d.git
2. pip install ./differentiable_t1d

Supported dynamics

• UVa/Padova 2008 simulator. See Kovatchev et al., 2009. In silico preclinical trials: a proof of concept in closed-loop control of type 1 diabetes.
• Medtronic Virtual Patient (MVP) model. See Kanderian et al., 2009. Identification of Intraday Metabolic Profiles during Closed-Loop Glucose Control in Individuals with Type 1 Diabetes