| Description |
Continuous glucose data were collected using a multimodal minimally invasive CGM system and clinical indicators such as demographics data and laboratory results were also collected. A clinical big dataset including numerous glucose data and corresponding clinical indicators were established. Through a cloud platform for in-depth mining of the above dataset, key signals for impending hypoglycemia were determined and thus used as inputs to establish a hypoglycemia risk prediction model. The model was further validated and optimized in a prospective cohort. Furthermore, multimodal physiological signals of minimally invasive CGM, diet type, exercise type, and intensity were collected. After using the deep learning method and functional CGM parameters, a personalized regulation model based on CGM was established to accurately guide patients' diet and exercise, which will be verified and optimized in a prospective cohort. |