Optimizing the ICA-based removal of ocular EEG artifacts from free viewing experiments

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Combining EEG and eye-tracking offers insights into neural correlates of natural vision.
• Ocular artifacts from eye movements heavily contaminate EEG recordings.
• Current artifact correction methods, like ICA, often leave residual artifacts under free viewing conditions.

Purpose of the Study:

• To systematically evaluate and optimize ICA-based artifact correction for EEG during unconstrained eye movements.
• To compare the performance of optimized ICA with standard settings and an alternative spatial filter (MSEC).
• To quantify correction quality by assessing both residual artifacts and removal of neurogenic activity.

Main Methods:

• Orthogonal variation of four ICA pipeline parameters: high-pass filter, low-pass filter, proportion of training data with saccadic spike potentials (SP), and eye-tracker-based component rejection threshold.
• Objective quantification of correction quality using eye-tracking data to measure under- and overcorrection.
• Comparison with Multiple Source Eye Correction (MSEC) as a benchmark.

Main Results:

• Commonly used ICA settings resulted in residual artifacts and distorted neurogenic activity.
• Optimizing ICA by training on filtered data with overweighted saccadic spike potentials (SPs) significantly improved artifact removal.
• Optimized ICA successfully removed virtually all artifacts, including SPs and their associated broadband artifacts, with minimal distortion of neural activity.
• Optimized ICA outperformed MSEC in correcting SPs.

Conclusions:

• Optimized ICA procedures provide effective artifact correction for EEG during natural vision tasks with free eye movements.
• The optimized method significantly reduces ocular artifacts, including saccadic spike potentials, without compromising neural signal integrity.
• This approach offers a more robust method for analyzing EEG data combined with eye-tracking.