Machine learning for the diagnosis of early-stage diabetes using temporal glucose profiles

摘要: Machine learning shows remarkable success for recognizing patterns in data. Here, we apply machine (ML) the diagnosis of early-stage diabetes, which is known as a challenging task medicine. Blood glucose levels are tightly regulated by two counter-regulatory hormones, insulin and glucagon, failure homeostasis leads to common metabolic disease, diabetes mellitus. It chronic disease that has long latent period complicates detection at an early stage. The vast majority cases result from diminished effectiveness action, resistance modifies temporal profile blood glucose. Thus, propose use ML detect subtle changes pattern concentration. Time series data on with sufficient resolution currently unavailable, so confirm proposal using synthetic profiles produced biophysical model considers regulation hormone action. Multi-layered perceptrons, convolutional neural networks, recurrent networks all identified degree high accuracy above $$85\%$$ .