Individualized short-term core temperature prediction in humans using biomathematical models.

Area of Science:

• Physiology
• Biomathematics
• Sports Science

Background:

• Accurate prediction of body core temperature during physical activity is crucial for preventing heat-related illnesses.
• Mathematical modeling offers a promising approach to personalize temperature monitoring.

Purpose of the Study:

• To compare the predictive accuracy of three mathematical modeling techniques for individual-specific body core temperature during physical activity.
• To evaluate the portability of these models across individuals and study conditions.

Main Methods:

• Comparison of a first-principles (SCENARIO) model, a purely data-driven model, and a hybrid model.
• Validation using two distinct datasets: a Field study and a Laboratory study.
• Assessment of prediction accuracy for a short-term horizon (20-30 minutes ahead).

Main Results:

• For a prediction horizon up to 30 minutes, the purely data-driven model demonstrated the highest accuracy, followed by the hybrid model.
• The first-principles SCENARIO model exhibited approximately twice the root mean square prediction error compared to the data-driven and hybrid models.
• Data-driven and hybrid models, when properly developed, showed potential for portability across individuals and studies.

Conclusions:

• Purely data-driven and hybrid models are superior to the first-principles SCENARIO model for short-term (≤30 min) body core temperature prediction during physical activity.
• The portability of data-driven and hybrid models can significantly reduce the effort required for individual-specific model development and data collection.