Sparse time artifact removal.

Area of Science:

• Electrophysiology
• Signal Processing
• Neuroscience

Background:

• Muscle artifacts and electrode noise impede electrophysiological signal interpretation.
• These artifacts are often channel-specific, limiting the effectiveness of techniques like Independent Component Analysis (ICA).
• High-frequency artifacts can obscure or mimic gamma band cortical activity.

Purpose of the Study:

• To introduce and evaluate the Sparse Time Artifact Removal (STAR) algorithm.
• To address the challenge of channel-specific artifacts in electrophysiological data.
• To improve the interpretability of electrophysiological signals by removing noise.

Main Methods:

• The STAR algorithm partitions data into artifact-free and contaminated segments.
• It estimates data correlation from the artifact-free portion's covariance matrix.
• Artifacts are corrected by projecting channels onto a subspace defined by other channels.

Main Results:

• STAR demonstrated high effectiveness in removing or reducing channel-specific artifacts in both simulated and real data.
• The algorithm minimizes data loss compared to trial or channel removal methods.
• Processing is localized in time, preserving most of the original data and full dimensionality.

Conclusions:

• STAR complements existing linear component analysis techniques like ICA.
• It enhances the discovery of weak neural sources by increasing the number of effective noise-free channels.
• STAR offers a valuable tool for improving the quality and interpretability of electrophysiological recordings.