Blood Glucose Level Prediction Using Optimized Neural Network For Virtual Patients

Diabetes mellitus has been increasing to become one of the major and global problems. Studies reveal that complications associated with diabetes can be reduced by the viable management of Blood Glucose Levels (BGLs). Objective: Two problems related to glucose-insulin regulation are discussed here. Firstly, the prior prediction of blood glucose to overcome the lag time for insulin absorption. Secondly, data recording issue for diabetes patients is always a big problem in the diagnosis of diabetes type 1 patients. Method: In this research, we input Continuous Glucose Monitoring (CGM) data to Optimized Artificial Neural Networks (OANN) in order to predict BGL of Type 1 Diabetes (T1D). We have investigated virtual CGM data of 2 subjects in order to depict the efficiency of the proposed method and to validate the ANN. These two case studies have been compiled from AIDA i.e. the freeware mathematical diabetes simulator. Results: For BGL predictions, improved results have been shown for minimal inputs in the prediction horizon (PH) of 15, 30, 45 and 60 min. Results produced by experimentation reveal that our ANN is not only accurate, but adaptive, and encouraging as well and thus can be implemented in clinics. Furthermore, this study targets to make life easier for T1D patients by minimizing human input to the system. Conclusion and Future work: In the future, we intend to investigate a greater collection of AIDA scenarios, and data that is real and influence other factors of BGLs.