Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrating Bayesian Approaches and Expert Knowledge for Forecasting Continuous Glucose Monitoring Values in Type 2 Diabetes Mellitus

Yuyang Sun, Panagiotis Kosmas

TL;DR

This study makes the first application of the ShanghaiT2DM dataset for glucose level forecasting, considering the influences of diabetes-related variables, and establishes a foundational framework for developing personalized diabetes management strategies, potentially enhancing diabetes care through more accurate and timely interventions.

Abstract

Precise and timely forecasting of blood glucose levels is essential for effective diabetes management. While extensive research has been conducted on Type 1 diabetes mellitus, Type 2 diabetes mellitus (T2DM) presents unique challenges due to its heterogeneity, underscoring the need for specialized blood glucose forecasting systems. This study introduces a novel blood glucose forecasting system, applied to a dataset of 100 patients from the ShanghaiT2DM study. Our study uniquely integrates knowledge-driven and data-driven approaches, leveraging expert knowledge to validate and interpret the relationships among diabetes-related variables and deploying the data-driven approach to provide accurate forecast blood glucose levels. The Bayesian network approach facilitates the analysis of dependencies among various diabetes-related variables, thus enabling the inference of continuous glucose monitoring (CGM) trajectories in similar individuals with T2DM. By incorporating past CGM data including inference CGM trajectories, dietary records, and individual-specific information, the Bayesian structural time series (BSTS) model effectively forecasts glucose levels across time intervals ranging from 15 to 60 minutes. Forecast results show a mean absolute error of 6.41 mg/dL, a root mean square error of 8.29 mg/dL, and a mean absolute percentage error of 5.28%, for a 15-minute prediction horizon. This study makes the first application of the ShanghaiT2DM dataset for glucose level forecasting, considering the influences of diabetes-related variables. Its findings establish a foundational framework for developing personalized diabetes management strategies, potentially enhancing diabetes care through more accurate and timely interventions.

Integrating Bayesian Approaches and Expert Knowledge for Forecasting Continuous Glucose Monitoring Values in Type 2 Diabetes Mellitus

TL;DR

This study makes the first application of the ShanghaiT2DM dataset for glucose level forecasting, considering the influences of diabetes-related variables, and establishes a foundational framework for developing personalized diabetes management strategies, potentially enhancing diabetes care through more accurate and timely interventions.

Abstract

Precise and timely forecasting of blood glucose levels is essential for effective diabetes management. While extensive research has been conducted on Type 1 diabetes mellitus, Type 2 diabetes mellitus (T2DM) presents unique challenges due to its heterogeneity, underscoring the need for specialized blood glucose forecasting systems. This study introduces a novel blood glucose forecasting system, applied to a dataset of 100 patients from the ShanghaiT2DM study. Our study uniquely integrates knowledge-driven and data-driven approaches, leveraging expert knowledge to validate and interpret the relationships among diabetes-related variables and deploying the data-driven approach to provide accurate forecast blood glucose levels. The Bayesian network approach facilitates the analysis of dependencies among various diabetes-related variables, thus enabling the inference of continuous glucose monitoring (CGM) trajectories in similar individuals with T2DM. By incorporating past CGM data including inference CGM trajectories, dietary records, and individual-specific information, the Bayesian structural time series (BSTS) model effectively forecasts glucose levels across time intervals ranging from 15 to 60 minutes. Forecast results show a mean absolute error of 6.41 mg/dL, a root mean square error of 8.29 mg/dL, and a mean absolute percentage error of 5.28%, for a 15-minute prediction horizon. This study makes the first application of the ShanghaiT2DM dataset for glucose level forecasting, considering the influences of diabetes-related variables. Its findings establish a foundational framework for developing personalized diabetes management strategies, potentially enhancing diabetes care through more accurate and timely interventions.
Paper Structure (17 sections, 6 equations, 5 figures, 7 tables, 1 algorithm)

This paper contains 17 sections, 6 equations, 5 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Description of Datasets
  4. Data Preprocessing
  5. Data-driven Bayesian Network Learning
  6. Knowledge-driven of Bayesian Network Structure Learning
  7. Bayesian Structural Time Series Forecasting
  8. Semi-local Liner Trend Component
  9. Seasonal Component
  10. Regressor Component
  11. Average Forecasting and MCMC
  12. Experiments and Results
  13. Experiments Settings
  14. Evaluation of Forecasting Performance on the ShanghaiT2DM Dataset: Results Across Various Prediction Horizons and Different Models
  15. Ablation Experiments of the BSTS Model Components
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Data-driven approach system architecture overview illustrating the integration of the Bayesian network and the BSTS models, accompanied by data flow representations. The 'DM clinical data flow' delineates the transmission of datasets encompassing anthropometric and biochemical characteristics related to both T1DM and T2DM. Concurrently, the 'DM time series data flow' illustrates the transmission of datasets that contain CGM measurements with dietary records. A detailed description of the Bayesian network is available in Section II-C and depicted in Figure \ref{['fig:fig_label_00']}.
  • Figure 2: The structure of 'Network_DM' resulting from structural Bayesian network learning on both ShanghaiT1DM and ShanghaiT2DM datasets. This network depicts 16 arcs: 7 consensus arcs (red arrows), 7 unique arcs (black arrows), and 2 potential arcs (brown dotted arrows). 'Network_DM' is an averaged representation based on $200$ structure learning models (both 'Model_1' and 'Model_2'), with each arc demonstrating a confidence level (arc strength) exceeding $0.85$. Nodes in the network correspond to characteristic features listed in Table \ref{['tab:my-table_1']}. Arc strengths for 'Network_DM' are detailed at the bottom of the chart.
  • Figure 3: Three-day period of CGM and forecasting trajectories for a Shanghai T2DM subject over a 15-minute prediction horizon, including 95% confidence interval upper and lower bounds of forecasts.
  • Figure 4: Three-day period of CGM and forecasting trajectories for a Shanghai T2DM subject over the 30-minute and 60-minute prediction horizon.
  • Figure 5: Confusion matrices for blood glucose forecasting results at (a) 15-minute, (b) 30-minute, and (c) 60-minute prediction horizons. Each matrix compares predicted glucose conditions (hypoglycemia, normal, hyperglycemia) against actual CGM measurements. The percentages indicate the proportion of all subjects' forecasting results falling into each category.