Large-scale calcium imaging reveals a systematic V4 map for encoding natural scenes.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual System Research

Background:

• Biological visual systems are adapted for processing natural scenes.
• Understanding visual cortical functions necessitates characterizing neuronal population encoding of natural images.

Purpose of the Study:

• To map the natural image preferences of neuronal populations in the V4 cortical area.
• To elucidate the organizational principles governing natural scene encoding in V4.

Main Methods:

• Utilized widefield calcium imaging to record V4 cortical responses to thousands of natural images in male macaques.
• Developed a deep-learning digital twin of V4 to map neural population preferences at a 100-µm scale.
• Validated model predictions with additional widefield and two-photon imaging, alongside feature attribution analysis.

Main Results:

• A detailed map of V4 revealed diverse functional domains, each encoding distinct natural image features.
• These domains represent a continuum from processing localized shape features to dispersed surface features.
• Identified organizing principles of natural scene encoding within the V4 cortex.

Conclusions:

• The study provides a high-resolution map of V4's functional organization for natural scenes.
• Deep learning models can effectively characterize neural population preferences and cortical organization.
• Findings advance our understanding of how the brain processes complex visual information.