Uncertainty-aware Blood Glucose Prediction from Continuous Glucose Monitoring Data
This work addresses the problem of improving real-time blood glucose prediction and adverse event identification for individuals with Type 1 diabetes by integrating principled uncertainty quantification.
This paper investigates uncertainty-aware neural network models for blood glucose prediction and adverse glycemic event identification in Type 1 diabetes. They found that Transformer-based models with evidential output heads provided the most effective framework, achieving higher predictive accuracies and better-calibrated uncertainty estimates that correlated with prediction errors.
In this work, we investigate uncertainty-aware neural network models for blood glucose prediction and adverse glycemic event identification in Type 1 diabetes. We consider three families of sequence models based on LSTM, GRU, and Transformer architectures, with uncertainty quantification enabled by either Monte Carlo dropout or through evidential output layers compatible with Deep Evidential Regression. Using the HUPA-UCM diabetes dataset for validation, we find that Transformer-based models equipped with evidential output heads provide the most effective uncertainty-aware framework, achieving consistently higher predictive accuracies and better-calibrated uncertainty estimates whose magnitudes significantly correlate with prediction errors. We further evaluate the clinical risk of each model using the recently proposed Diabetes Technology Society error grid, with risk categories defined by international expert consensus. Our results demonstrate the value of integrating principled uncertainty quantification into real-time machine-learning-based blood glucose prediction systems.