Dual-feature selectivity enables bidirectional coding in visual cortical neurons

Visual cortical neurons possess a unique encoding strategy that allows them to process multiple visual features simultaneously, according to a recent study. Traditionally, sensory neurons were thought of as individual feature detectors, responding to specific stimuli while remaining indifferent to others. However, this research reveals a dual-feature encoding mechanism in macaque visual cortex, as well as in mouse cortex, which challenges this view.

Many neurons in areas V1 and V4 were found to be selectively tuned to two distinct visual features. One feature enhances activity, while the other suppresses it, around an elevated baseline firing rate. This dual-feature encoding strategy was identified through a combination of neuronal recordings and functional digital twin models—deep learning-based predictive models of biological neurons. These models allowed researchers to systematically determine each neuron's preferred and non-preferred features.

These feature pairs acted as anchors for a continuous, low-dimensional axis in natural image similarity space. Along this axis, neuronal activity varied approximately linearly, suggesting a shared feature selectivity across the population. This shared selectivity across neurons within a visual area indicates that the encoding strategy is organized across the population, structuring stimulus encoding.

The encoding strategy of dual-feature selectivity was found to be conserved across species. It was observed in both primary and lateral visual areas of mouse cortex, further supporting its significance in visual processing. This strategy aligns with recent anatomical evidence for feature-specific inhibitory connectivity. Alongside this anatomical finding, the dual-feature encoding mechanism highlights the role of selective excitation and inhibition in enhancing the representational capacity of the neuronal population through circuit mechanisms.