Search Results for « Blood glucose prediction »

Accurate estimations for the near future levels of blood glucose are crucial for Type 1 Diabetes Mellitus (T1DM) patients in order to be able to react on time and avoid hypo and hyper-glycemic episodes. Accurate predictions for blood glucose are the base for control algorithms in glucose regulating systems such as the artificial pancreas. Numerous research studies have already been conducted in order to provide predictions for blood glucose levels with particularities in the input signals and underlying models used. These models can be categorized into two major families: those based on tuning glucose physiological-metabolic models and those based on learning glucose evolution patterns based on machine learning techniques. This paper reviews the state of the art in blood glucose predictions for T1DM patients and proposes, implements, validates and compares a new hybrid model that decomposes a deep machine learning model in order to mimic the metabolic behavior of physiological blood glucose methods. The differential equations for carbohydrate and insulin absorption in physiological models are modeled using a Recurrent Neural Network (RNN) implemented using Long Short-Term Memory (LSTM) cells. The results show Root Mean Square Error (RMSE) values under 5 mg/dL for simulated patients and under 10 mg/dL for real patients.

Abstract Background Accurate short-term blood glucose prediction can help reduce adverse glycemic events and improve diabetes management. Although machine learning models have shown promising performance for glucose forecasting, their clinical adoption remains limited due to a lack of transparency. This study aimed to evaluate both the predictive performance and interpretability of machine learning models for 30-minute blood glucose prediction in individuals with Type 1 and Type 2 diabetes. Methods We use continuous glucose monitoring and clinical data from Type 1 and Type 2 diabetes cohorts to develop five blood glucose prediction models: Linear Regression, Support Vector Regression, Random Forest, Extreme Gradient Boosting, and Long Short-Term Memory. Model performance was assessed using Root Mean Squared Error and the coefficient of determination. After that, explainable artificial intelligence (XAI) methods including SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) were applied to analyze global feature importance and explain individual predictions. Furthermore, comparative analyses were conducted across diabetes types and prediction scenarios. Results Support Vector Regression and tree-based ensemble models achieved the strongest predictive performance across both diabetes populations, while Long Short-Term Memory networks were constrained by limited data availability. Recent glucose measurements were consistently identified as the most influential predictors. Interpretability analysis revealed distinct glucose dynamics between diabetes types, with Type 1 diabetes exhibiting greater short-term volatility and Type 2 diabetes showing more stable patterns with additional contributions from clinical variables. SHAP produced more stable and comprehensive explanations, whereas LIME provided simpler but less robust local interpretations. Conclusions Machine learning models can provide accurate short-term glucose forecasts while maintaining meaningful interpretability through explainable artificial intelligence techniques. SHAP and LIME offer complementary strengths, with the former providing higher fidelity and the latter improving accessibility. These findings support the development of transparent and trustworthy artificial intelligence systems for diabetes management and clinical decision support.

Abstract This research presents a new evolving neural network approach to forecast blood glucose for people with diabetes. The accuracy of forecasting using the proposed evolving neural network is demonstrated to outperform a conventional back propagation neural network. People with diabetes need to control their blood sugar levels. High blood sugar over long term leads to many other health complications. To avoid high blood sugar, it is important for people to be able to predict what will happen to blood sugar so that they can do something to prevent hypo or hyper glycaemia. However, many external uncontrollable factors can make blood glucose difficult to predict, such as meals which increase carbs and glucose goes up. Exercise also affects blood glucose, but exercise can be aerobic or anaerobic and these affect blood glucose in opposite ways. There has been research aiming to predict blood glucose by analysing previous recorded data from continuous glucose monitoring devices. This research applies a new approach with evolutionary computation to evolve a neural network, using neuro evolution, and the optimised neural network is then applied to predict and forecast blood glucose changes. In the comparison of accuracy, the results show that evolved neural network outperformed a back-propagation neural network in this task on forecasting CGM data. This can help people with diabetes to have a better idea about how their blood glucose is going to change before it occurs, so that hypo and hyper can be avoided. This can reduce diabetes complications and costs for the health service.

Diabetes is a metabolic disorder characterized by persistent hyperglycemia, with its incidence steadily rising worldwide. Blood glucose monitoring is a core measure in diabetes management, and continuous glucose monitoring provides more comprehensive and accurate glucose data compared to traditional fingerstick testing. To collect continuous glucose data from patients, precise glucose prediction algorithms can help them better control their blood glucose fluctuations. Therefore, by addressing the issues of low prediction accuracy, complex input features, and poor generalization performance in existing glucose prediction methods, this paper proposes a glucose prediction model based on a double-layer SCINet stack using time-series analysis methods. SCINet effectively captures multi-scale dynamic features in time-series data through recursive down-sampling and convolution operations, making it suitable for glucose prediction tasks. Experimental data were sourced from real-world continuous glucose monitoring records of patients at Yixing People’s Hospital. Model input features were optimized through variable selection and data preprocessing, with predictive performance validated on a test dataset. The results demonstrate that the proposed model outperforms existing time-series prediction models across varying prediction horizons and patient datasets, exhibiting high predictive accuracy and stability.