Classifying Stress From Heart Rate Variability Using Salivary Biomarkers as Reference.

Area of Science:

• Physiological monitoring
• Biomarker analysis
• Machine learning applications

Background:

• Noninvasive stress assessment from human physiology is challenging.
• Existing methods often lack accuracy or require invasive procedures.
• Understanding the correlation between physiological signals and stress biomarkers is crucial.

Purpose of the Study:

• To propose a pattern recognition system for noninvasive stress assessment.
• To correlate heart rate variability (HRV) features with salivary stress biomarkers.
• To develop an accurate predictive model for stress responses.

Main Methods:

• Utilized the Trier Social Stress Test to induce stress in volunteers.
• Collected heart rate and salivary measurements (alpha-amylase, cortisol).
• Employed fuzzy ARTMAP (FAM) for correlating HRV features with biomarkers, enhanced by genetic algorithms and voting ensembles.

Main Results:

• Developed an ensemble of FAMs for predicting stress biomarkers from heart rate data.
• Achieved 75% accuracy in classifying stress using salivary alpha-amylase.
• Reached 80% accuracy when using salivary cortisol as the stress indicator.

Conclusions:

• The proposed pattern recognition system demonstrates effective noninvasive stress assessment.
• HRV features can be reliably used to predict stress levels indicated by salivary biomarkers.
• This approach offers a promising method for objective stress monitoring.