Mid-level visual features engage V4 to support sparse object recognition under uncertainty

Cortical visual prostheses have long promised to restore sight by directly stimulating the visual cortex, but current approaches remain limited. Conventional V1-targeted implants evoke isolated phosphenes that rarely support coherent object recognition, largely due to the sparse electrode arrays feasible for implantation. Here we examined whether mid-level features, specifically curve segments associated with area V4, provide a more efficient substrate for artificial vision. Using fMRI, 24 participants performed an object recognition task with stimuli fragmented into either phosphene-like dots or curve segments at graded densities. Behaviorally, curve segments enabled recognition at lower density thresholds than phosphenes, yielding steeper psychometric slopes and higher accuracy. Neuroimaging revealed that while both fragment types engaged V1, only curve segments recruited V4, and correct recognition of segments relative to phosphenes elicited widespread responses across temporal, parietal, and insular cortices. Mediation analyses further showed that recognition from phosphenes relied on anterior cingulate cortex (ACC) involvement with higher stimulus uncertainty, whereas segment-based recognition depended primarily on visual cortex. Together, these results demonstrate that mid-level feature fragments confer both perceptual and neural advantages over phosphenes, reducing reliance on downstream uncertainty resolution. V4 thus emerges as a promising target for next-generation cortical prostheses, offering a path toward sparser yet more informative stimulation strategies.