Salivary Markers for Quantitative Dehydration Estimation During Physical Exercise.

Area of Science:

• Exercise Physiology
• Biochemistry
• Machine Learning Applications

Background:

• Salivary markers are proposed as noninvasive indicators of dehydration during physical exercise.
• Previous threshold-based methods struggle with high intersubject variability, limiting accuracy.
• Existing research often assumes linear relationships between markers and dehydration.

Purpose of the Study:

• To develop a machine-learning approach for quantitative estimation of total body water (TBW) loss.
• To address and overcome intersubject variabilities in salivary dehydration markers.
• To explore novel salivary markers beyond traditional ones.

Main Methods:

• Collected salivary samples and TBW loss data from ten subjects during a 2-hour running workout.
• Analyzed salivary osmolality, proteins, amylase, chloride, cortisol, cortisone, and potassium.
• Applied a Gaussian process machine-learning model for quantitative TBW loss estimation.

Main Results:

• Quantitative TBW loss estimation achieved with an error of approximately 0.34 liters.
• Demonstrated nonlinear increases in salivary markers during progressive dehydration.
• Identified potential for improved physiological models based on nonlinear marker behavior.

Conclusions:

• Machine learning effectively handles intersubject variability for accurate dehydration assessment.
• Salivary markers exhibit nonlinear dynamics with increasing dehydration, challenging prior linear assumptions.
• This approach enables more precise total body water loss estimations, aiding hydration management.