Due to the sensitive nature of diabetes-related data, preventing them from shared between studies, progress in the field of glucose prediction is to assess. To address this issue, we present GLYFE (GLYcemia Forecasting), a benchmark of machine-learning-based glucose-predictive models. To ensure the reproducibility of the results and the usability of the in the future, we provide extensive details about the data flow. Two are used, the first comprising 10 in-silico adults from the UVA/Padova 1 Diabetes Metabolic Simulator (T1DMS) and the second being made of 6 real-1 diabetic patients coming from the OhioT1DM dataset. The predictive are personalized to the patient and evaluated on 3 different prediction (30, 60, and 120 minutes) with metrics assessing their accuracy and acceptability. The results of nine different models coming from the glucose-prediction are presented. First, they show that standard autoregressive linear are outclassed by kernel-based non-linear ones and neural networks. In, the support vector regression model stands out, being at the same one of the most accurate and clinically acceptable model. Finally, the performances of the models are the same for both datasets. This shows, even though data simulated by T1DMS are not fully representative of-world data, they can be used to assess the forecasting ability of the-predictive models. Those results serve as a basis of comparison for future studies. In a field data are hard to obtain, and where the comparison of results from studies is often irrelevant, GLYFE gives the opportunity of gathering around a standardized common environment.
No takes yet. Share an insight, caveat, or question.
Bois et al. (2020) studied this question.