Inference-enabled tracking of acute mental stress via multi-modal wearable physiological sensing: A proof-of-concept study

Objective

To develop a novel algorithm for tracking acute mental stress which can infer acute mental stress state from multi-modal digital signatures of physiological parameters compatible with wearable-enabled sensing.

Methods

We derived prominent digital signatures of physiological responses to mental stress using cross-integration of multi-modal physiological signals including the electrocardiogram (ECG), photoplethysmogram (PPG), seismocardiogram (SCG), ballistocardiogram (BCG), electrodermal activity (EDA), and respiratory effort. Then, we developed an algorithm for tracking acute mental stress that can continuously classify stress vs no stress states by computing an aggregated likelihood computed with respect to a priori probability density distributions associated with the digital signatures of mental stress under stress and no stress states.

Results

Our algorithm could adequately infer mental stress state (average classification accuracy: 0.85, sensitivity: 0.85, specificity: 0.86) using a small number of prominent digital signatures derived from cross-integration of multi-modal physiological signals. The digital signatures in our work significantly outperformed the digital signatures employed in the state-of-the-art in tracking acute mental stress. Its exploitation of collective inference allowed for improved inference of mental stress state relative to naïve data mining techniques.

Conclusion

Our algorithm for tracking acute mental stress has the potential to make a leap in continuous, high-accuracy, and high-confidence inference of mental stress via convenient wearable-enabled physiological sensing. Significance: The ability to continuously monitor and track mental stress can collectively improve human wellbeing.