Can AI Help Manage Type 1 Diabetes?

Diabetes impacts millions of people in the United States alone.

According to the Centers for Disease Control and Prevention, type 1 diabetes affects 5-10% of people with diabetes. Those with type 1 diabetes don’t produce enough insulin on their own, so they require insulin from outside sources.

Managing type 1 diabetes can be taxing for individuals and their families. As technology and medicine progress, we’re seeing more and more ways that people can check their blood sugar (like by using a smartphone app) or even treat diabetes with new medicine.

With the advancements in artificial intelligence, people are looking to find ways that it can help in medical settings. Harry Emerson, a PhD student at the University of Bristol’s Department of Engineering Mathematics, is studying to see how AI can help people living with type 1 diabetes.

In a study published in June in the Journal of Biomedical Informatics, Emerson partnered with the University Hospital Southampton. In the study, Emerson worked to teach a machine-learning algorithm to keep virtual diabetes patients alive.

According to the study, titled, “Offline reinforcement learning for safer blood glucose control in people with type 1 diabetes,” the researchers trained the AI using seven months’ worth of data from 30 simulated patients. The AI was able to learn how much insulin to deliver in a variety of real-life scenarios. It managed to calculate a dosing strategy on par with commercial controllers.

The study only tested virtual patients, as real-life testing would prove risky. After all, insulin dosing is a life-or-death situation for patients with type 1 diabetes.

In the end, the AI insulin monitor outperformed current state-of-the-art controllers in the simulation. In fact, the AI system kept virtual patients in their target blood glucose range an hour longer per day when compared with commercial controllers.

Since the simulation test went so well, Emerson hopes to move testing to real people.