Advanced time blood glucose value prediction

Increased glucose level measurement data with the introduction of continuous glucose monitoring (CGM), a new technology in the field of diabetes; accelerated the application of artificial intelligence methods in the prediction of advanced time blood glucose value. Advanced time blood glucose value prediction, that is, predicting the glucose values in the blood before a certain period of time, enables diabetic patients to better manage their glucose levels in the blood, to receive early warnings about the wrong treatments and unwanted conditions such as hypoglycemia hyperglycemia. In the study, a new physiological model was designed to model the effect of insulin and carbohydrate data on glucose value. Plasma insulin concentration and glucose absorption rate data created based on the personalized parameter values obtained by optimizing various parameters in the newly designed physiological model are used in the artificial neural network and long short term memory network together with the past CGM data, personalized blood glucose value prediction was made at the 30 and 60 minute prediction horizon. According to the filtering processes on CGM data and the method used, 4 different models were created and the results of these models were compared and presented. It has been clearly shown that filtering operations on CGM data decrease the estimation performance, increase the estimation performance of the data calculated with personal parameters, and for this problem, the artificial neural network method with short processing time and less complexity gives better results than the long short term memory method.