Sparse Multichannel Decomposition of Electrodermal Activity With Physiological Priors.

Area of Science:

• Physiology
• Biomedical Engineering
• Signal Processing

Background:

• Inferring autonomous nervous system (ANS) activity is crucial for understanding stress regulation.
• Skin conductance (SC) variations reflect sudomotor nerve activity (SMNA) and ANS-driven physiological responses.
• Single-channel SC analysis is limited by noise and artifacts, hindering accurate ANS activity inference.

Purpose of the Study:

• To develop a robust method for inferring ANS activity from multichannel SC data.
• To improve the accuracy of recovering ANS activations and physiological system parameters.
• To overcome limitations of single-channel approaches in the presence of measurement noise.

Main Methods:

• Modeling skin conductance with a second-order differential equation and cubic basis splines.
• Utilizing a block coordinate descent method for SC signal decomposition.
• Employing generalized cross-validation and physiological priors for sparse recovery.

Main Results:

• The proposed algorithm accurately recovers ANS activations, tonic and phasic components, and system parameters.
• Multichannel analysis significantly outperforms single-channel methods in noise performance and signal reconstruction.
• Demonstrated superior performance across multiple metrics including AUC score and goodness of fit.

Conclusions:

• Concurrent decomposition and deconvolution of multichannel SC signals offer significant advantages for ANS activity inference.
• The developed algorithm provides a more accurate and reliable method for analyzing SC data.
• This approach has critical applications in stress regulation and understanding physiological responses.