ospEDA: Orthogonal Subspace Projection for Electrodermal Activity Decomposition

Area of Science:

• Physiological signal processing
• Biomedical engineering
• Neuroscience

Background:

• Electrodermal activity (EDA) is crucial for measuring sympathetic nervous system responses like stress and arousal.
• Accurate decomposition of EDA into tonic and phasic components is difficult, especially with noisy data and varied experimental conditions.

Purpose of the Study:

• To introduce ospEDA, a novel Orthogonal Subspace Projection (OSP)-based method for robust EDA decomposition.
• To evaluate ospEDA's performance against existing methods using simulated and real-world datasets.

Main Methods:

• Physiologically informed valley detection for tonic component estimation.
• Orthogonal Subspace Projection (OSP) for phasic component extraction, accommodating inter-subject variability.
• Non-negative least squares (NNLS) deconvolution with ridge regularization for phasic driver estimation.

Main Results:

• ospEDA demonstrated competitive performance in simulations, with low RMSE for tonic (0.131) and phasic (0.132) components at 20 dB SNR.
• Maintained strong phasic component accuracy (RMSE=0.293, r=0.782, R²=0.979) under noisier conditions (10 dB SNR).
• Achieved highest F1-scores for sympathetic nerve activity detection across various SNRs and showed significant baseline vs. stimulus separation (AUROC=0.765) in real-world data.

Conclusions:

• ospEDA offers a robust and consistent framework for EDA decomposition, particularly effective in varying noise levels.
• The method provides reliable phasic driver estimation, enhancing its utility for physiological monitoring applications.
• ospEDA shows promise for advancing the analysis of sympathetic nervous system activity in diverse research and clinical settings.