Systematic Review on Deep Learning Algorithms for Blood Glucose Forecasting in Type 1 Diabetes.

Area of Science:

• Endocrinology and Metabolism
• Artificial Intelligence in Medicine
• Biomedical Data Science

Background:

• Type 1 Diabetes (T1D) management relies on continuous glucose monitoring (CGM) for real-time blood glucose (BG) data.
• Forecasting algorithms, particularly deep learning (DL), leverage CGM data to predict future BG levels, aiding therapeutic interventions.
• A comprehensive review of DL applications for BG prediction is needed to guide clinical adoption.

Purpose of the Study:

• To systematically review the current state of deep learning applications for blood glucose prediction in Type 1 Diabetes.
• To evaluate DL models based on dataset characteristics, inputs, training, architecture, and performance metrics.
• To identify challenges and future research directions for DL-based BG forecasting.

Main Methods:

• Systematic literature review following PRISMA guidelines.
• Searches conducted across PubMed, Scopus, and Web of Science databases.
• Analysis of 26 selected studies focusing on DL models for BG prediction in T1D.

Main Results:

• Deep learning models demonstrate significant potential for accurate BG forecasting using CGM data.
• Evaluated studies varied in dataset characteristics, model inputs, architectures, and prediction horizons.
• Key challenges include ensuring physiological fidelity and interpretability of DL models for clinical use.

Conclusions:

• Deep learning models offer a promising approach for real-time BG prediction in T1D management.
• Explainable AI (XAI) integration is crucial for enhancing model reliability, safety, and clinical adoption.
• Future research should focus on developing interpretable and physiologically sound DL models for T1D care.