Abstract:In the UK, approximately 400,000 people with type 1 diabetes (T1D) rely on insulin delivery due to insufficient pancreatic insulin production. Managing blood glucose (BG) levels is crucial, with continuous glucose monitoring (CGM) playing a key role. CGM, tracking BG every 5 minutes, enables effective blood glucose level prediction (BGLP) by considering factors like carbohydrate intake and insulin delivery. Recent research has focused on developing sequential models for BGLP using historical BG data, incorporating additional attributes such as carbohydrate intake, insulin delivery, and time. These methods have shown notable success in BGLP, with some providing temporal explanations. However, they often lack clear correlations between attributes and their impact on BGLP. Additionally, some methods raise privacy concerns by aggregating participant data to learn population patterns. Addressing these limitations, we introduced a graph attentive memory (GAM) model, combining a graph attention network (GAT) with a gated recurrent unit (GRU). GAT applies graph attention to model attribute correlations, offering transparent, dynamic attribute relationships. Attention weights dynamically gauge attribute significance over time. To ensure privacy, we employed federated learning (FL), facilitating secure population pattern analysis. Our method was validated using the OhioT1DM'18 and OhioT1DM'20 datasets from 12 participants, focusing on 6 key attributes. We demonstrated our model's stability and effectiveness through hyperparameter impact analysis.

What problem does this paper attempt to address?

The paper aims to address the issue of blood glucose level prediction for patients with Type 1 Diabetes (T1D). Specifically: - **Research Background**: Approximately 400,000 people in the UK suffer from Type 1 Diabetes. Since their pancreas cannot produce enough insulin, they need to rely on insulin pumps to maintain normal blood glucose levels. Continuous Glucose Monitors (CGM) record blood glucose levels every 5 minutes. This data can be used to predict future blood glucose levels, helping patients take appropriate preventive measures (such as insulin injections or carbohydrate intake) to avoid various complications caused by unstable blood glucose levels. - **Problems with Existing Methods**: Current blood glucose level prediction methods mainly focus on building complex time-series models to capture dynamic features in historical blood glucose time series and consider additional attributes (such as carbohydrate intake, insulin injections, etc.). Although these methods perform well in blood glucose level prediction and some have temporal interpretability, they fail to clearly show the correlation and importance of each input attribute. Additionally, some methods learn group patterns by merging data from all participants, which infringes on individual privacy. - **Proposed Method**: To overcome the above limitations, the authors propose a Graph Attention Memory (GAM) model based on Graph Attention Networks (GAT) and Gated Recurrent Units (GRU). GAT uses graph attention mechanisms to model the relationships between multiple attributes and presents these transparent relationships through dynamic graph structures. GRU is used to aggregate the graph features extracted by GAT, generating compact tensors for blood glucose level prediction. Furthermore, the paper introduces Federated Learning (FL) to protect participants' privacy while learning group patterns. - **Experimental Results**: The model was evaluated on the OhioT1DM'18 and OhioT1DM'20 datasets. When predicting blood glucose levels 30 minutes and 60 minutes ahead, the Root Mean Square Error (RMSE) was 18.753 mg/dL and 31.275 mg/dL, respectively, and the Mean Absolute Relative Difference (MARD) was 8.893% and 15.903%, respectively. Compared to the top eight methods in the 2020 blood glucose level prediction challenge, this model performed excellently. Additionally, the paper visualized the dynamic correlations between various activity attributes over different time periods.

Blood Glucose Level Prediction: A Graph-based Explainable Method with Federated Learning

Privacy Preserved Blood Glucose Level Cross-Prediction: An Asynchronous Decentralized Federated Learning Approach

GARNN: An Interpretable Graph Attentive Recurrent Neural Network for Predicting Blood Glucose Levels via Multivariate Time Series

FedGlu: A personalized federated learning-based glucose forecasting algorithm for improved performance in glycemic excursion regions

Dynamic Partitioning of Graphs Based on Multivariate Blood Glucose Data—A Graph Neural Network Model for Diabetes Prediction

Deep Multitask Learning by Stacked Long Short-Term Memory for Predicting Personalized Blood Glucose Concentration

Optimization and Evaluation of an Intelligent Short-Term Blood Glucose Prediction Model Based on Noninvasive Monitoring and Deep Learning Techniques

Deep transfer learning: a novel glucose prediction framework for new subjects with type 2 diabetes

Blood Glucose Monitoring Using Non-Invasive Features of Wearable Devices and Machine Learning

Blood Glucose Level Forecasting on Type-1-Diabetes Subjects during Physical Activity: A Comparative Analysis of Different Learning Techniques

Machine Learning based prediction of Glucose Levels in Type 1 Diabetes Patients with the use of Continuous Glucose Monitoring Data

Blood Glucose Prediction with Variance Estimation Using Recurrent Neural Networks

Long-term Prediction of Blood Glucose Levels in Type 1 Diabetes Using a CNN-LSTM-Based Deep Neural Network

Blood glucose forecasting from temporal and static information in children with T1D

Exploring Nutritional Influence on Blood Glucose Forecasting for Type 1 Diabetes Using Explainable AI

Population-Specific Glucose Prediction in Diabetes Care With Transformer-Based Deep Learning on the Edge

Deep transfer learning and data augmentation improve glucose levels prediction in type 2 diabetes patients

Feature Transformation for Efficient Blood Glucose Prediction in Type 1 Diabetes Mellitus Patients

The Challenge of Predicting Meal-to-meal Blood Glucose Concentrations for Patients with Type I Diabetes

Multi-Horizon Glucose Prediction Across Populations with Deep Domain Generalization

A State-by-State online transfer learning framework with incremental clustering for blood glucose prediction