Movie Features Are Hierarchically Encoded in the Dynamics of Functional Connectivity.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Visual perception involves hierarchical processing in the cortex.
• Convolutional neural networks (CNNs) can extract features from natural stimuli.
• Previous studies linked stimulus complexity to visual stream activation but not functional connectivity dynamics.

Purpose of the Study:

• To investigate how specific stimulus features influence functional brain connectivity during naturalistic viewing.
• To compare the temporal dynamics of CNN feature extraction with brain activity.
• To understand the spectral and topological organization of functional connectivity related to stimulus complexity.

Main Methods:

• Acquired Magnetoencephalography (MEG) data from 12 subjects watching movies.
• Utilized CNNs to track dynamic movie features.
• Compared temporal dynamics of CNN layers with functional connectivity in alpha and beta bands.

Main Results:

• Functional connectivity dynamics aligned with stimulus temporal evolution in alpha and beta bands.
• Increased stimulus complexity led to more connections and longer temporal delays.
• Higher stimulus complexity shifted network hub involvement from low-level to high-level visual regions.
• Alpha band correlated with mid-level features; beta band correlated with higher-level features.

Conclusions:

• Stimulus features are encoded in the temporal, spectral, and topological organization of functional brain connectivity.
• Brain communication dynamics reflect the complexity and hierarchy of visual input.
• Distinct spectral bands (alpha and beta) play specialized roles in processing different levels of visual information.