CGMformer: a novel deep-learning model promising for early detection of prediabetes to effectively prevent type 2 diabetes

Type 2 diabetes (T2D) is a chronic disorder in glucose metabolic homeostasis.Its crucial molecular pathological mechanism is, driven by both genetic and environmental factors, obstruction of insulin signal transduction pathway leading to characteristic insulin resistance and -cell dysfunction [1-3].Despite ongoing efforts in lifestyle intervention, the global incidence of T2D continues to rise annually [4].To seize the time-window of reversible stage of this disorder and to effectively reduce its incidence, early detection is an important starting step.Therefore, it is very imperative to have a relative accurate yet convenient measure to detect its early stage prediabetes.A recent interdisciplinary study--conducted by a mathematics team led by Yong Wang and Luonan Chen, and a medical team led by Weiping Jia and Huating Li--developed a deep-learning model named CGMformer, meeting such demand [5].CGMformer was pretrained on large-scale and diverse corpus of Continuous Glucose Monitoring (CGM) data from the Nationwide Multicenter CGM datasets and unlabeled National Real-World CGM data.CGM is used for monitoring blood glucose on a continual basis with a device consisting of a small electrode probe sensor placed under the skin, an attached transmitter, and a separate receiver that displays glucose levels.