A real-time algorithm for predicting core temperature in humans.

Area of Science:

• Physiology
• Biomedical Engineering
• Data Science

Background:

• Accurate core body temperature monitoring is crucial for heat illness prevention.
• Existing offline algorithms require further development for real-time application.
• Real-time prediction can enable timely interventions during heat exposure.

Purpose of the Study:

• To implement and evaluate a real-time algorithm for human core temperature prediction.
• To assess the accuracy and uncertainty of real-time core temperature forecasts.
• To validate the algorithm's utility as a heat illness warning system.

Main Methods:

• Developed a real-time algorithm using a zero-phase Butterworth digital filter and an autoregressive (AR) model.
• Simulated real-time performance using core temperature data from two field studies (10 individuals).
• Evaluated prediction accuracy via Root Mean Squared Error (RMSE) and uncertainty via Prediction Intervals (PIs).

Main Results:

• Real-time predictions showed slightly higher RMSE than offline methods but remained practically accurate for ~20-minute horizons.
• The algorithm successfully alerted to critical temperature thresholds in a heat-ill individual.
• Prediction intervals accurately reflected forecast uncertainty based on prediction horizon and temperature variability.

Conclusions:

• The real-time algorithm is a feasible and effective tool for predicting human core temperature.
• It demonstrates potential as a warning system for impending heat illnesses.
• The study supports the feasibility of 'universal' AR models applicable across individuals.