Blood Glucose Prediction in Type 1 Diabetes Based on Long Short-Term Memory

The management of Type 1 Diabetes (T1D) has always been a complex task, requiring patients to continuously monitor their blood glucose levels and adjust insulin doses accordingly. This paper explores the potential of Artificial Intelligence (AI), specifically Long Short-Term Memory (LSTM) networks, to revolutionize the way Type 1 Diabetes is managed. This research takes a novel approach by focusing solely on blood glucose levels, aiming to determine if Artificial Intelligence can effectively predict and recommend insulin doses based on this singular data point. In this paper, we compare three variants of Long short-term memory (LSTM) models, with the purpose of blood glucose level prediction: unidirectional LTSM, stacked LTSM and Bi-directional LTSM. The data used for this study is unique in that it is sourced from a single individual, ensuring consistency and eliminating inter-individual variability. The findings presented in this paper could serve as a foundation for future research and the development of Artificial Intelligence driven diabetes management systems.