Visual feature processing in a large stroke cohort: evidence against modular organization.

Abstract

Mid-level visual processing represents a crucial stage between basic sensory input and higher-level object recognition. The conventional model posits that fundamental visual qualities, such as colour and motion, are processed in specialized, retinotopic brain regions (e.g. V4 for colour, MT/V5 for motion). Using atlas-based lesion-symptom mapping and disconnectome maps in a cohort of 307 ischaemic stroke patients, we examined the neuroanatomical correlates underlying the processing of eight mid-level visual qualities. Contrary to the predictions of the standard model, our results did not reveal consistent relationships between processing impairments and damage to traditionally associated brain regions. Although we validated our methodology by confirming the established relationship between visual field defects and damage to primary visual areas (V1, V2 and V3), we found no reliable evidence linking processing deficits to specific regions in the posterior brain. These findings challenge the traditional modular view of visual processing and suggest that mid-level visual processing might be more distributed across neural networks than previously thought. This supports alternative models where visual maps represent constellations of co-occurring information rather than specific qualities.