tDAI: Type 1 Diabetes Closed-loop AI

Problem:

Type 1 Diabetes is a rapidly growing issue, estimates are as high as 8.4 million people worldwide with over 400,000 diagnoses each year according to The Lancet. Recently, closed loop systems have taken an interest from pharmaceutical companies due to their effectiveness in lowering HBA1c levels by continuously monitoring glucose levels using Continuous Glucose Monitors (CGM's) and adjusting insulin dosages in real-time. However, most systems do not take into account a persons unique glucose trends and insulin sensitivity, which can hinder their effectiveness by not accounting for these factors and could potentially be life-threatening/fatal in some cases.

Purpose:

This project's purpose is to research how Artificial Intelligence accounts for these factors, its effectiveness and reliability, and how this system could be implemented on top of preexisting closed-loop systems or individually.

Plan of action:

We plan on initially recreating glucose trends within TensorFlow from a donor's CGM data, which will help us create testcases for our interstitial fluid models, mock-up insulin pumps, and glucose injectors so we can develop and test a closed-loop system using Keras, whilst comparing its accuracy to preexisting models, such as the t:slim X2 CLS.