GluNet: A Deep Learning Framework for Accurate Glucose Forecasting
GluNet: A Deep Learning Framework for Accurate Glucose Forecasting
复制标题
DOI：
10.1109/jbhi.2019.2931842
复制
发表时间：
2020-02-01
影响因子：
7.7
通讯作者：
Georgiou, Pantelis
中科院分区：
文献类型：
--
作者：
Li, Kezhi;Liu, Chengyuan;Georgiou, Pantelis
对于1型糖尿病（T1 D）患者，血糖（BG）预测可用于有效避免高血糖、低血糖和相关并发症。最新的连续葡萄糖监测（CGM）技术允许人们实时观察葡萄糖。然而，准确的葡萄糖预测仍然是一个挑战。在这项工作中，我们引入了GluNet，这是一个利用个性化深度神经网络的框架，可以根据T1 D受试者的历史数据（包括血糖测量值、膳食信息、胰岛素剂量和其他因素）预测T1 D受试者短期（30-60分钟）未来CGM测量值的概率分布。它采用了最新的深度学习技术，包括四个部分：数据预处理，标签转换/恢复，多层扩张卷积神经网络（CNN）和后处理。该方法在成人和青少年受试者中进行了计算机模拟评价。结果表明，显着改善现有的方法在文献中，通过全面的比较，在均方根误差（RMSE）（8.88 +/- 0.77 mg/dL），时滞较短（0.83 +/- 0.40分钟），对于预测层位（PH）= 30分钟（分钟），对于虚拟成人受试者，PH = 60分钟的RMSE（19.90 +/- 3.17 mg/dL）和时滞（16.43 +/- 4.07 min）。此外，GluNet还在两个临床数据集上进行了测试。结果显示，对于PH = 30分钟，其实现了RMSE（19.28 +/- 2.76 mg/dL）和时滞（8.03 +/- 4.07分钟），对于PH = 60分钟，其实现了RMSE（31.83 +/- 3.49 mg/dL）和时滞（17.78 +/- 8.00分钟）。与其他方法相比，这些是葡萄糖预测的最佳报告结果，包括用于预测葡萄糖的神经网络（NNPG），支持向量回归（SVR），具有外源输入的潜在变量（LVX）和具有外源输入的自回归（ARX）算法。
For people with Type 1 diabetes (T1D), forecasting of blood glucose (BG) can be used to effectively avoid hyperglycemia, hypoglycemia and associated complications. The latest continuous glucosemonitoring (CGM) technology allows people to observe glucose in real-time. However, an accurate glucose forecast remains a challenge. In this work, we introduce GluNet, a framework that leverages on a personalized deep neural network to predict the probabilistic distribution of short-term (30-60 minutes) future CGM measurements for subjects with T1D based on their historical data including glucose measurements, meal information, insulin doses, and other factors. It adopts the latest deep learning techniques consisting of four components: data pre-processing, label transform/recover, multilayers of dilated convolution neural network (CNN), and post-processing. The method is evaluated in-silico for both adult and adolescent subjects. The results show significant improvements over existing methods in the literature through a comprehensive comparison in terms of root mean square error (RMSE) (8.88 +/- 0.77 mg/dL) with short time lag (0.83 +/- 0.40 minutes) for prediction horizons (PH) = 30 mins (minutes), and RMSE (19.90 +/- 3.17 mg/dL) with time lag (16.43 +/- 4.07 mins) for PH = 60 mins for virtual adult subjects. In addition, GluNet is also tested on two clinical data sets. Results show that it achieves an RMSE (19.28 +/- 2.76 mg/dL) with time lag (8.03 +/- 4.07 mins) for PH = 30 mins and an RMSE (31.83 +/- 3.49 mg/dL) with time lag (17.78 +/- 8.00 mins) for PH = 60 mins. These are the best reported results for glucose forecasting when compared with other methods including the neural network for predicting glucose (NNPG), the support vector regression (SVR), the latent variable with exogenous input (LVX), and the auto regression with exogenous input (ARX) algorithm.