Decoding state specific connectivity during speech production and perception

Area of Science:

• Cognitive Neuroscience
• Systems Neuroscience
• Neuroimaging

Background:

• Understanding dynamic brain networks is crucial for language processing.
• Functional connectivity (FC) using fMRI has limitations, especially in speech production.
• Electrocorticography (ECoG) offers a high-resolution alternative for studying brain networks.

Purpose of the Study:

• To investigate brain networks supporting speech perception and production using ECoG.
• To decode cognitive states from functional connectivity patterns during language tasks.
• To determine the contribution of both active and less active brain regions to cognitive states.

Main Methods:

• Utilized ECoG data from 42 patients during controlled speech production tasks.
• Acquired neural activity across five cognitive states: auditory, picture, and reading perception, speech production, and baseline.
• Employed linear classifiers to decode cognitive states from single-trial FC (Pearson correlations).

Main Results:

• Achieved a mean decoding accuracy of 64.4% for cognitive states from FC patterns.
• Identified distinct network signatures for auditory and visual perception, and speech production.
• Revealed that regions with minimal neural activation are critical for differentiating cognitive states.

Conclusions:

• Functional connectivity analysis provides complementary insights beyond local neural activity.
• Dynamic brain networks, including less active regions, are essential for language perception and production.
• ECoG is a valuable tool for dissecting the neural underpinnings of complex cognitive functions like speech.