Semantic attributes are encoded in human electrocorticographic signals during visual object recognition.

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Non-invasive neuroimaging demonstrates semantic category and attribute encoding in neural activity.
• Electrocorticography (ECoG) offers advantages, but its capacity for encoding semantic attributes is unexplored.

Purpose of the Study:

• To investigate the extent to which semantic attribute information is encoded in ECoG responses.
• To develop and validate high-dimensional encoding models for mapping semantic attributes to neural activity.

Main Methods:

• Recorded ECoG data from patients naming objects across 12 semantic categories.
• Trained high-dimensional encoding models to correlate semantic attributes with spectral-temporal ECoG features.
• Utilized encoding models to decode untrained objects and analyze neural correlates of semantic dimensions.

Main Results:

• Decoding accuracy for untrained objects using ECoG models was comparable to whole-brain functional Magnetic Resonance Imaging (fMRI).
• High-gamma activity (70-110Hz) in basal occipitotemporal electrodes correlated with specific semantic dimensions (manmade-animate, size, places-tools).
• Individual patient results aligned with established findings on semantic processing along the ventral visual pathway.

Conclusions:

• The semantic attribute encoding model approach is effective for decoding objects not included in the training set.
• ECoG can capture complex semantic encodings, offering a powerful tool for studying object recognition.
• Findings support the utility of ECoG for understanding the neural basis of semantic representation.