Managing diabetes can be challenging, especially when it comes to keeping blood sugar levels under control after meals. However, advancements in technology such as the integration of artificial intelligence (AI) in diabetes management devices have brought about new possibilities for improved glycemic control. A recent study from Oregon Health & Science University investigated the potential benefits of a robust insulin delivery system that utilizes AI for meal detection and carbohydrate content estimation. The system, known as the Robust Artificial Pancreas (RAP), uses a neural network model to automatically detect meals and deliver a recommended insulin dose. This randomized, single-center crossover trial compared postprandial glucose control in the four hours following unannounced meals using a hybrid model predictive control (MPC) algorithm and the RAP system. The meal detection algorithm has a sensitivity of 83.3%, false discovery rate of 16.6%, and mean detection time of 25.9 minutes.

      The study found that the RAP system significantly reduces time spent with high blood sugar levels and trends towards increasing time spent in the optimal blood sugar range (70-180 mg/dL) by 9.1% compared to traditional meal planning methods. No significant difference was observed in the Time below range (glucose <70 mg/dL) between RAP and MPC. The results of the study show that using AI-powered insulin delivery systems can be an effective way to manage blood sugar levels after meals. The RAP system has shown that if a meal is accurately detected within 25-30 minutes of the meal occurring and dosed a percentage of the required insulin, time spent with high blood sugar levels can be significantly reduced with no increased risk of hypoglycemia.

      According to Mosqueria-Lopez, the lead investigator, this research demonstrates that automated insulin delivery systems can be enhanced by using machine-learning-based models for meal detection and carbohydrate content calculation. This approach can also be applied in metabolic simulators to design and assess tools prior to clinical studies, providing more support for the use of these tools in diabetes management.

Read More