Can AI predict glucose changes before they happen?

Diabetes affects a significant portion of the adult population in Europe, and those with type 1 diabetes require constant monitoring and careful insulin management. While glucose levels can be influenced by various factors, predicting fluctuations in glucose levels could help assess risks and improve diabetes management. Researchers at Kaunas University of Technology (KTU) have developed an artificial intelligence (AI) model that forecasts glucose changes and assesses the risk of hypoglycemia up to 30 and 60 minutes ahead.

This model combines glucose readings with information on insulin delivery, carbohydrate intake, and physical activity, representing each data point as a node in a graph network and connecting it to similar points in the past. The researchers used an attention mechanism to identify which earlier data points are most relevant to the outcome, similar to an experienced doctor assessing a patient's condition.

The model was tested using international type 1 diabetes datasets and showed high accuracy in both glucose forecasting and hypoglycemia risk assessment. When trained to perform both tasks simultaneously, the algorithm performed even better. However, the personalized insulin adjustments evaluated in the study were not intended to allow patients to alter their treatment independently but to serve as a potential decision-support tool for clinical analysis.

While the technology shows promise, further clinical studies are needed before it can be widely adopted in clinical practice.