Interpreting Deep Glucose Predictive Models for Diabetic People Using RETAIN

TL;DR

This paper evaluates the RETAIN deep learning model for glucose prediction in diabetics, demonstrating its interpretability and comparable accuracy to LSTM models, with potential benefits for patient and clinician understanding.

Contribution

It introduces the use of the RETAIN architecture for glucose forecasting, highlighting its interpretability and effectiveness compared to other models.

Findings

RETAIN outperforms random forest in accuracy metrics.

RETAIN matches LSTM performance on clinical metrics.

Tools developed enhance interpretability for patients and practitioners.

Abstract

Progress in the biomedical field through the use of deep learning is hindered by the lack of interpretability of the models. In this paper, we study the RETAIN architecture for the forecasting of future glucose values for diabetic people. Thanks to its two-level attention mechanism, the RETAIN model is interpretable while remaining as efficient as standard neural networks. We evaluate the model on a real-world type-2 diabetic population and we compare it to a random forest model and a LSTM-based recurrent neural network. Our results show that the RETAIN model outperforms the former and equals the latter on common accuracy metrics and clinical acceptability metrics, thereby proving its legitimacy in the context of glucose level forecasting. Furthermore, we propose tools to take advantage of the RETAIN interpretable nature. As informative for the patients as for the practitioners, it can enhance the understanding of the predictions made by the model and improve the design of future glucose predictive models.