Personalized blood glucose models for exercise, meal and insulin interventions in type 1 diabetic children

Area of Science:

• Biomedical Engineering
• Endocrinology
• Computational Biology

Background:

• Modern healthcare trends emphasize personalized, predictive, preventive, and participatory (P4) medicine to enhance patient quality of life (QoL).
• Accurate blood glucose (BG) prediction models that integrate lifestyle interventions are crucial for preventing hypoglycemia and diabetes complications in type 1 diabetes (T1D).

Purpose of the Study:

• To develop personalized BG prediction models for T1D children by incorporating lifestyle interventions.
• To utilize time series modeling to create patient-specific BG forecasting tools.

Main Methods:

• Employed multi-input single-output time series models to develop personalized BG models for 12 T1D children.
• Utilized clinical data including rate of perceived exertion (RPE), carbohydrate absorption, and insulin kinetics as model inputs.
• Applied linear models (Box-Jenkins, state space, process transfer function) and nonlinear Hammerstein-Wiener models.

Main Results:

• Successfully developed personalized BG models for all 12 T1D children, capturing inter-patient variability.
• Linear models were suitable for 9 patients, while nonlinear Hammerstein-Wiener models were optimal for the remaining 3.
• Demonstrated the feasibility of using diverse modeling approaches to create individualized BG prediction systems.

Conclusions:

• Personalized BG prediction models integrating lifestyle factors are achievable for T1D children.
• The developed models can aid in creating tailored exercise, diet, and insulin prescriptions.
• This approach holds potential for improving diabetes management and preventing adverse events.