Uncertainty quantification in neural network-based glucose prediction for diabetes

TL;DR

This study evaluates uncertainty-aware neural network models, including LSTM, GRU, and Transformer architectures, for blood glucose prediction in Type 1 diabetes, emphasizing improved accuracy and uncertainty calibration.

Contribution

It introduces the use of evidential output layers with Transformer models for uncertainty quantification in glucose prediction, demonstrating superior performance.

Findings

Transformer models with evidential outputs outperform others in accuracy.

Uncertainty estimates correlate strongly with prediction errors.

Models show improved clinical risk assessment using the Diabetes Technology Society error grid.

Abstract

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.