A data driven approach to understanding the organization of high-level visual cortex.

Area of Science:

• Neuroscience
• Cognitive Science
• Visual Perception

Background:

• Scene-selective regions in the human visual cortex, like the PPA, are associated with image semantics and categories.
• The role of fundamental organizing principles in these neural representations remains unclear.
• Previous studies often used experimenter-defined stimuli, potentially limiting insights into basic visual dimensions.

Purpose of the Study:

• To investigate whether neural responses in scene-selective areas reflect fundamental visual organizing principles.
• To explore the contribution of basic image properties versus semantic information in shaping neural representations.
• To provide an image-based explanation for higher-level visual representations.

Main Methods:

• A data-driven approach was used to analyze a large database of scene images (>100,000).
• Images were described by visual properties: orientation, spatial frequency, and spatial location.
• K-means clustering identified distinct regions in the visual feature space, and selected images from these clusters.

Main Results:

• Clusters of images, not corresponding to typical scene categories, elicited distinct neural response patterns in the PPA.
• The similarity of neural responses in the PPA correlated with the similarity of image properties across clusters.
• Neural responses were predicted by perceptual judgments of scenes but not by their semantic content.

Conclusions:

• The findings suggest that the PPA's organization is based on fundamental visual features rather than semantic categories.
• This study offers an image-based framework for understanding how higher-level representations emerge in scene-selective brain regions.
• Basic visual properties play a crucial role in shaping neural representations of scenes.