Research of neural network-based blood glucose level forecasting systems for insulin-dependant diabetes patients
暂无分享,去创建一个
This paper documents the results of the research involving neural network-based blood glucose level forecasting systems for insulin-dependent diabetes patients. Forecast is made for continuous subcutaneous insulin injections and continuous subcutaneous glucose measurements. Elman, layer-recurrent, and NARX network architectures were considered in the research. The influence of the network architecture, the number of neurons, the training algorithm and the tapped delay line on the forecast precision were investigated. The research is part of the optimal insulin dose determination algorithm creation work, which can complement the insulin pump and continuous glucose monitoring system to a closed-loop system that will perform artificial pancreas functions. The research resulted in the determination of the optimal neural network architecture that allowed obtaining the necessary forecast precision for the short-range predictions.
[1] Hans-Jörg Pfleiderer,et al. Modelling the glucose metabolism with backpropagation through time trained Elman nets , 2003, 2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718).
[2] Yang Kuang,et al. Mathematical models and software tools for the glucose-insulin regulatory system and diabetes: an overview , 2006 .
[3] P. Wach,et al. Neural predictive controller for insulin delivery using the subcutaneous route , 1998, IEEE Transactions on Biomedical Engineering.