Dual-feature selectivity enables bidirectional coding in visual cortical neurons

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Cortex Research

Background:

• Traditional view: Sensory neurons act as simple feature detectors, increasing firing rate for preferred stimuli.
• Limited understanding of complex neural coding strategies beyond basic excitation.
• Need to explore how neurons encode multiple stimulus properties simultaneously.

Purpose of the Study:

• To identify and characterize a novel dual-feature encoding strategy in the visual cortex.
• To investigate the role of this strategy in neural population coding of visual information.
• To explore the cross-species and cross-area conservation of this encoding mechanism.

Main Methods:

• Neuronal recordings in macaque visual areas V1 and V4.
• Development and application of functional digital twin models (deep learning-based predictive models).
• Analysis of neuronal activity along continuous axes in natural image similarity space.

Main Results:

• Discovery of neurons selectively tuned to two distinct visual features: one enhancing and one suppressing activity.
• Identification of a continuous, low-dimensional axis in image space where neuronal activity varies approximately linearly.
• Evidence of shared feature selectivity across neuronal populations, structuring stimulus encoding.
• Conservation of dual-feature selectivity across species (macaque and mouse) and visual areas (V1, V4, primary, and lateral areas).

Conclusions:

• Neurons employ a sophisticated dual-feature encoding strategy involving both excitation and suppression.
• This strategy increases the representational capacity of neuronal populations.
• Findings align with anatomical evidence for feature-specific inhibitory connectivity.
• Dual-feature selectivity represents a fundamental coding principle in the visual cortex.