A review of continuous blood glucose monitoring and prediction of blood glucose level for diabetes type 1 patient in different prediction horizons (PH) using artificial neural network (ANN)

Abstract

Prior blood-glucose level prediction is necessary for management of diabetes therapy. Continuous Glucose Monitoring (CGM) data can be used to predict future blood glucose levels in order to control hypoglycemia/hyperglycemic events. The forecasted blood glucose level can be feed into Artificial Pancreas (AP) for an autonomous smart glucose-insulin regulation. In this field, neural network (NN) has proved its efficiency, performance, and reliability. In this paper, we are aiming to find gaps and further improvements in the domain. We review different forms of neural networks, different size of datasets, features selected, real and virtual datasets and different prediction horizon (PH) from 15 to 75 min. We filtered 15 research papers between 2010 and 2018 for blood glucose level prediction which are using Artificial Neural Networks. This paper provides brief of each+ study and how it is contributing to this domain. It also highlights the advantages of each research study and how they can be improved to get high accuracy and precision for blood glucose level prediction without decreasing the prediction horizon. This study is helpful in opening a gateway for new researchers to identify the future work and to carry out their research in that direction.