Quantifying Signal Quality From Unimodal and Multimodal Sources: Application to EEG With Ocular and Motion Artifacts

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Increasing use of portable, non-invasive devices for electrophysiological data collection.
• Data quality is crucial for the effectiveness of medical devices relying on these signals.
• Challenges exist in assessing the quality of real-world electrophysiological recordings.

Purpose of the Study:

• To develop and validate novel methods for evaluating electrophysiological signal quality.
• To quantify the extent to which recorded data represent the physiological source of interest.
• To assess the performance of artifact removal algorithms.

Main Methods:

• Investigated data-driven models, including Bayesian decision (unimodal) and deep learning (multimodal) approaches.
• Scored electroencephalography (EEG) data quality based on ocular and motion artifacts.
• Validated methods on three EEG datasets (N=60 subjects).

Main Results:

• Demonstrated effective scoring of EEG data quality using both unimodal and multimodal methods.
• Successfully applied the unimodal method to compare the performance of two artifact removal algorithms.
• Validated the ability to assess data quality and evaluate noise-reduction algorithms.

Conclusions:

• The developed methods provide a reliable way to score electrophysiological data quality.
• These methods can be used to objectively evaluate the effectiveness of artifact removal techniques.
• The approach supports the reliable use of real-world electrophysiological data in medical applications.