The importance of artificial intelligence in postprandial blood glucose prediction for insulin bolus calculation

• Author(s): Miranda, Francisco ; Amorim, Débora ; Ferreira, Luís ; Abreu, Carlos
• Date: 2024
• Persistent ID: http://hdl.handle.net/10773/42091
• Origin: RIA - Repositório Institucional da Universidade de Aveiro
• Document Type: Journal article
• Language: English

Description

Many works are using artificial intelligence to forecast postprandial blood glucose. However, the following questions arise: is it necessary to develop artificial intelligence techniques to predict blood glucose? How important is artificial intelligence for this purpose? This work gives some insights seeking the answer to these questions in the context of using postprandial blood glucose predictions to optimize the prandial insulin bolus. Considering the bolus optimization model proposed in this work, the error in the postprandial glycemia due to an inaccurate postprandial blood glucose prediction is in the same amount as the error made in the prediction. Therefore, more accurate postprandial blood glucose predictions lead to postprandial blood glucose values closer to the predefined blood glucose target defined for that patient. In this way, it is possible to conclude that artificial intelligence could have a relevant role in helping patients control their blood glucose levels. In particular, regarding non-controlled patients with high glucose variability.