Network Communications Flexibly Predict Visual Contents That Enhance Representations for Faster Visual Categorization

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Visual Perception

Background:

• Visual cognition models propose predictive brain networks facilitate stimulus categorization.
• Understanding network-level prediction and categorization from neural signals is challenging.
• Existing methods struggle to isolate specific information processing pathways.

Purpose of the Study:

• To reconstruct and analyze brain network mechanisms for predicting and categorizing visual stimuli.
• To investigate how content-specific communications within brain networks influence behavior.
• To differentiate predictive network activity from general neural communication.

Main Methods:

• Used magnetoencephalography (MEG) to record neural signals from participants (N=11).
• Applied novel connectivity measures to isolate content-specific network communications.
• Reconstructed prediction and categorization networks for low (LSF) vs. high (HSF) spatial frequency stimuli.

Main Results:

• Identified a top-down Prediction Network (temporal to occipital cortex) controlled by the prefrontal cortex.
• Demonstrated that predictions enhance bottom-up sensory representations in an occipital-ventral-parietal-frontal Categorization Network.
• Showed that isolated content communications represent a subset (55-75%) of overall neural communication.

Conclusions:

• The study successfully isolated functional networks underlying cognitive functions like prediction and categorization.
• Findings support a model where top-down predictions interact with bottom-up sensory input for perception.
• The identified networks and their dynamic interactions provide new insights into cognitive information processing in the brain.