New preprint: DeepRetinotopy toolbox
This work introduces the DeepRetinotopy toolbox, a general, dataset-agnostic application that enables the prediction of functional topographic maps of the human visual cortex from cortical anatomy alone — without the need to acquire functional brain data. The toolbox requires only a standard anatomical scan (a T1-weighted MRI image), which represents a significant advance in the general applicability of deep learning techniques to predicting topographic organization. We have comprehensively validated the toolbox for generalisability across diverse acquisition protocols and demonstrate how it can be readily applied to existing large-scale datasets, using over 11,000 anatomical brain scans. This provides a strong framework for accurate, individual-specific prediction of functional topographies that govern the functional organisation of the visual cortex, and the brain more generally.