eduzhai > Applied Sciences > Engineering >

GLYFE Review and Benchmark of Personalized Glucose Predictive Models in Type-1 Diabetes

  • Save

... pages left unread,continue reading

Document pages: 15 pages

Abstract: Due to the sensitive nature of diabetes-related data, preventing them frombeing shared between studies, progress in the field of glucose prediction ishard to assess. To address this issue, we present GLYFE (GLYcemia ForecastingEvaluation), a benchmark of machine-learning-based glucose-predictive models.To ensure the reproducibility of the results and the usability of thebenchmark in the future, we provide extensive details about the data flow. Twodatasets are used, the first comprising 10 in-silico adults from the UVA PadovaType 1 Diabetes Metabolic Simulator (T1DMS) and the second being made of 6 realtype-1 diabetic patients coming from the OhioT1DM dataset. The predictivemodels are personalized to the patient and evaluated on 3 different predictionhorizons (30, 60, and 120 minutes) with metrics assessing their accuracy andclinical acceptability.The results of nine different models coming from the glucose-predictionliterature are presented. First, they show that standard autoregressive linearmodels are outclassed by kernel-based non-linear ones and neural networks. Inparticular, the support vector regression model stands out, being at the sametime one of the most accurate and clinically acceptable model. Finally, therelative performances of the models are the same for both datasets. This showsthat, even though data simulated by T1DMS are not fully representative ofreal-world data, they can be used to assess the forecasting ability of theglucose-predictive models.Those results serve as a basis of comparison for future studies. In a fieldwhere data are hard to obtain, and where the comparison of results fromdifferent studies is often irrelevant, GLYFE gives the opportunity of gatheringresearchers around a standardized common environment.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...