Predicting blood glucose using an LSTM Neural Network

Abstract

Diabetes self-management relies on the blood glucose prediction as it allows taking suitable actions to prevent low or high blood glucose level. In this paper, we propose a deep learning neural network (NN) model for blood glucose prediction. It is a sequential one using a Long-Short-Term Memory (LSTM) layer with two fully connected layers. Several experiments were carried out over data of 10 diabetic patients to decide on the model's parameters in order to identify the best variant of it. The performance of the proposed LSTM NN measured in terms of root mean square error (RMSE) was compared with the ones of an existing LSTM and an autoregressive (AR) models. The results show that our LSTM NN is significantly more accurate; in fact, it outperforms the existing LSTM model for all patients and outperforms the AR model in 9 over 10 patients, besides, the performance differences were assessed by the Wilcoxon statistical test. Furthermore, the mean of the RMSE of our model was 12.38 mg/dl while it was 28.84 mg/dl and 50.69 mg/dl for AR and the existing LSTM respectively.

Forecasting of Glucose Levels and Hypoglycemic Events: Head-to-Head Comparison of Linear and Nonlinear Data-Driven Algorithms Based on Continuous Glucose Monitoring Data Only

“…Three ML models, successfully used in a wide range of regression problems, were considered: support vector regression (SVR) [ 39 , 40 ], regression random forest (RegRF) [ 41 ], and feed forward neural network (fNN) [ 42 ]. In addition, we considered a DL model, namely, long short-term memory (LSTM) network, which has shown promising results in glucose prediction [ 43 , 44 ]. The key idea of the SVR model is to map CGM data into a higher-dimensional feature space via a nonlinear mapping and, then, to perform a linear regression in such space [ 45 ].…”

“…Concerning LSTM, given the dimensions of our dataset and the elevated number of hyperparameters to be tuned, we decided to manually set some of them, such as the number of layers, learning rate, and decay factor, on the basis of literature studies to avoid the risk of overfitting [ 44 , 51 ]. This approach proved to be efficient in reducing such a risk in even more complex and deep neural networks [ 15 , 16 , 21 ].…”

Deep Physiological Model for Blood Glucose Prediction in T1DM Patients

“…The results outperformed previous methods, achieving RMSE values of 21.7 mg/dL and 36.9 mg/dL for prediction horizons of 30 and 60 min, respectively. A similar study using an LSTM RNN to predict upcoming values for BG levels can be found in [ 23 ]. The mean value of the RMSE of the model was 12.38 mg/dL based on data from 10 children and only used previous BG levels to estimate upcoming values.…”

Machine Learning Based Diabetes Classification and Prediction for Healthcare Applications

“…As future work, we plan to implement the android application for the proposed hypothetical diabetic monitoring system with the proposed classification and prediction approaches. Genetic algorithms can also be explored with the proposed prediction mechanism for better monitoring [24,64,[66][67][68][69][70][71].…”

“…e proposed approach outperformed as compared to other state-of-the-art techniques implemented, as shown in Table 2. LSTM is based on recurrent neural network (RNN) architecture, and it has feedback connections that make it suitable for diabetes forecasting [58]. LSTM mainly consists of a cell, keep gate, write gate, and an output gate, as shown in Figure 3.…”