Functional bipartite invariance in mouse primary visual cortex receptive fields

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual System Research

Background:

• Sensory systems generalize by identifying persistent features despite input variations.
• Understanding the neural basis of these invariances is challenging due to complex neural computations.

Purpose of the Study:

• To characterize neuronal invariances in mouse primary visual cortex (V1) using a scalable methodology.
• To investigate the role of neuronal invariances in visual processing and segmentation.

Main Methods:

• Utilized the inception loop paradigm: large-scale recordings, predictive modeling, and in silico experiments with in vivo verification.
• Synthesized varied exciting inputs (VEIs) to probe neuronal responses.
• Analyzed the MICrONS dataset for hierarchical analysis of neuronal invariances.

Main Results:

• Discovered a bipartite invariance in V1: one subfield encodes shift-tolerant high-frequency texture, another encodes a fixed low-frequency pattern.
• This division correlates with object boundaries defined by spatial frequency differences, suggesting a role in visual segmentation.
• Found a hierarchy in V1 layers 2/3 where postsynaptic neurons show greater invariance than presynaptic inputs, and less invariant neurons form more connections.

Conclusions:

• The study provides insights into the neuronal basis of invariances in the primary visual cortex.
• The developed inception loop paradigm offers a scalable methodology for mapping neuronal invariances.
• The findings suggest a contribution of V1 neuronal invariances to visual segmentation and object recognition.