딥러닝 기반 제2형 당뇨인의 식후 혈당 반응 예측: 연속혈당계와 장내 미생물 데이터 활용

Abstract

While carbohydrates are viewed as key post-prandial glycemic response (PPGR) predictors for type 2 diabetes (T2D) patients, their reliability has been debated. We hypothesized that it would be feasible to develop a personalized PPGR prediction model designed for patients with T2D, by incorporating additional personal features such as the clinical, medication, and relevant microbial features. Machine-learning techniques have improved PPGR predictions in healthy individuals; however, data for T2D patients remains sparse. Here, we developed personalized PPGR prediction algorithms with continuous glucose monitoring (CGM), gut microbiome, clinical and medication data in T2D patients using deep learning approach. Our best model obtained 0.704 correlation between predicted and actual PPGR. This investigation clearly suggests that deep learning model including CGM and personal features can significantly enhance prediction rates, paving the pathway for the real-world medical application of PPGR predictor in T2D individuals.