Bio-Inspired Strategies for Modeling and Detection in Diabetes Mellitus Treatment

6: Multistep predictor applied to T1DM patients

Abstract

For the treatment of diabetes mellitus, real-time monitoring devices have been developed to provide glucose measurements, these observations can be used for online glucose predictors to prevent risk scenarios such as hypoglycemia and hyperglycemia. Currently, deep learning for medical decision support systems has positively impacted the quality of life in society through the development of intelligent technologies. Deep neural networks are proposed for glucose prediction, these are: Long-term memory, Bidirectional long-term memory, and Multilayer Perceptron. The prediction horizons used are 15, 30, 45, and 60 min, which provide the glucose trend at various steps in the future that can help ...