A practical deep learning model for core temperature prediction of specialized workers in high-temperature...

Area of Science:

• Occupational Health and Safety
• Biomedical Engineering
• Environmental Health

Background:

• Rising global temperatures increase the frequency and severity of heat exposure during hazardous operations.
• Elevated core body temperature (Tcr) impairs physiological function and poses significant health risks to workers.
• Real-time Tcr monitoring is crucial but challenging in extreme heat environments.

Purpose of the Study:

• To develop a non-invasive model for accurate, real-time prediction of core body temperature (Tcr).
• To provide a practical tool for health monitoring and protection of personnel in high-temperature work environments.
• To address the limitations of current Tcr monitoring methods in hazardous conditions.

Main Methods:

• A novel non-invasive prediction model integrating a Kalman filter with a deep learning long-term sequence forecasting model.
• Utilized monitored skin temperature (Tsk) and heart rate (HR) as input features for personalized Tcr predictions.
• Validated the model through chamber experiments with participants under controlled high-temperature conditions (34-40°C).

Main Results:

• The model achieved high accuracy, with Mean Absolute Error (MAE) of 0.07, Root Mean Square Error (RMSE) of 0.09, and R-squared (R²) of 0.93.
• 95% of all Tcr predictions fell within an error margin of ±0.17°C.
• The model demonstrated effectiveness using seven-point Tsk combined with HR data.

Conclusions:

• The developed non-invasive Tcr prediction model offers a practical and accurate solution for health monitoring in extreme heat.
• Simple input requirements and high predictive accuracy make it suitable for protecting workers in hazardous, high-temperature operations.
• This technology can significantly enhance occupational safety and mitigate heat-related health risks.