Removing speech artifacts from electroencephalographic recordings during overt picture naming

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Speech Production

Background:

• Electroencephalography (EEG) studies examine brain activity during word production.
• Articulatory movements can obscure cognitive processes in EEG data.
• Accurate analysis requires effective removal of speech-related artifacts.

Purpose of the Study:

• To determine the timing and impact of speech artifacts in word production.
• To develop and evaluate a novel method for speech artifact removal (SAR).
• To compare the new SAR method with existing techniques.

Main Methods:

• Recorded EEG and lip electromyograms (EMGs) during picture naming.
• Utilized a picture-word interference paradigm.
• Compared a new SAR method based on Independent Component Analysis (SAR-ICA) with prior methods.

Main Results:

• Speech artifacts were detected earlier than previously thought, starting around 400 ms post-stimulus onset.
• The SAR-ICA method significantly reduced speech artifacts, showing weak correlation with lip EMG.
• EEG data corrected with SAR-ICA revealed neural activation patterns consistent with established word production models.

Conclusions:

• Speech artifacts significantly impact EEG studies of word production, occurring earlier than commonly assumed.
• The novel SAR-ICA method offers superior artifact attenuation compared to existing techniques.
• Accurate artifact removal is crucial for reliable investigation of cognitive processes in speech production.