BibTex format
@inproceedings{Zhu:2018,
author = {Zhu, T and Li, K and Herrero, P and Chen, J and Georgiou, P},
pages = {64--78},
title = {A deep learning algorithm for personalized blood glucose prediction},
year = {2018}
}
Publications
Citation
RIS format (EndNote, RefMan)
TY - CPAPER
AB - A convolutional neural network (CNN) model is presented to forecast the future glucose levels of the patients with type 1 diabetes. The model is a modified version of a recently proposed model called WaveNet, which becomes very useful in acoustic signal processing. By transferring the task into a classification problem, the model is mainly built by casual dilated CNN layers and employs fast WaveNet algorithms. The OhioT1DM dataset is the source of the four input fields: glucose levels, insulin events, carbohydrate intake and time index. The data is fed into the network along with the targets of the glucose change in 30 minutes. Several pre-processing approaches such as interpolation, combination and filtering are used to fill up the missing data in the training sets, and they improve the performance. Finally, we obtain the predictions of the testing dataset and evaluate the results by the root mean squared error (RMSE). The mean value of the best RMSE of six patients is 21.72.
AU - Zhu,T
AU - Li,K
AU - Herrero,P
AU - Chen,J
AU - Georgiou,P
EP - 78
PY - 2018///
SN - 1613-0073
SP - 64
TI - A deep learning algorithm for personalized blood glucose prediction
ER -