Neural correlates of local parallelism during naturalistic vision

Area of Science:

• Neuroscience
• Cognitive Science
• Computer Vision

Background:

• Human visual perception rapidly processes complex scenes by grouping elements into meaningful units.
• Neural mechanisms of perceptual grouping are poorly understood, especially using real-world stimuli.
• Local parallelism, a cue for grouping, has not been extensively studied in natural scene perception.

Purpose of the Study:

• To uncover the neural mechanisms of local parallelism in real-world scene perception.
• To determine the role of local parallelism in scene categorization.
• To identify the brain regions involved in processing local parallelism.

Main Methods:

• Developed an algorithm to detect local symmetry in images.
• Manipulated local parallelism in real-world scenes (line drawings and photographs).
• Used functional magnetic resonance imaging (fMRI) to record brain activity in 38 human observers.
• Performed model-based fMRI analysis on the BOLD5000 dataset.

Main Results:

• Scene categories were decoded more accurately from brain activity when scenes had strong local parallelism compared to weak parallelism, particularly in the parahippocampal place area (PPA).
• A strong relationship between local symmetry and brain activity was found in visual areas V1-V4, PPA, and retrosplenial cortex (RSC).
• The parallelism-related signal peaked first in visual area V4, suggesting it as an early processing site.

Conclusions:

• Local parallelism is a key perceptual grouping cue influencing neuronal activity across the visual hierarchy, starting in V4.
• Parallelism plays a significant role in how the PPA represents scene categories.
• This research provides insights into the neural basis of natural scene perception and visual grouping.