Application field
Diabetology, Artificial Pancreas (AP)
Technology
Artificial Intelligence (AI), Data Analysis, Closed-loop control systems
Goals
The Artificial Pancreas, combining autonomous insulin delivery and blood glucose monitoring, is a promising solution to treat Type 1 Diabetes (T1D). However, current AP systems still have some important limitations, such as the control of postprandial glucose and the regulation of insulin delivery at mealtimes.
Main goals:
- Study of the influence of nutritional factors on blood glucose levels up to 180 minutes after the meal.
- Development of a fully automated AI-based system for blood glucose prediction, able to handle postprandial glucose variations.
Activities
- Data analysis and pre-processing pipeline on real data from several T1D patients.
- Design and implementation of a supervised machine learning framework (Feed Forward Neural Network) for the prediction of blood glucose levels after 15, 30, 45, and 60 minutes from the meal leveraging on insulin doses, blood glucose history, and nutritional factors.
Challenges
- Intra-subjects and inter-subjects variability in blood glucose trends and insulin metabolism.
- Unpredictability of blood glucose levels when the prediction horizon is extended at several hours after the meal.
- Investigation of the input-output relationships considering eXplainable Artificial Intelligence (XAI) methods.
Bibliography
- Angrisani, L., Annuzzi, G., Arpaia, P., Bozzetto, L., Cataldo, A., Corrado, A., De Benedetto, E., Di Capua, V., Prevete, R., and Vallefuoco, E. (2022, May). Neural Network-Based Prediction and Monitoring of Blood Glucose Response to Nutritional Factors in Type-1 Diabetes. In 2022 IEEE International Instrumentation and Measurement Technology Conference (I2MTC) (pp. 1-6).