Article ID Journal Published Year Pages File Type
4961688 Procedia Computer Science 2016 7 Pages PDF
Abstract

It is naturally desirable to avoid hypo/hyperglycemic events andcommercial devices exist that can alert the patient before they occur. It is known, however, that percentage of false alerts for those devices is still high and much is still needed to be done to improve that.The purpose of this paper is to design a blood glucose prediction system that can be used aspart of a continuous glucose monitoring (CGM) device. With the help of a Kalman filter, glucose concentration is first reducedof its random noise component, and a neural network is then used for prediction of glucose upto two hours. Finally, this system is thoroughly tested for accuracy against various externalfactors. It is shown that such factors as patient's body weight, his/her exercise period andlifestyle may influence how well glucose concentration is predicted and therefore should betaken into account for early and accurate detection of hypo/hyperglycemic episodes.

Related Topics
Physical Sciences and Engineering Computer Science Computer Science (General)
Authors
, ,