Self-Organized Retinotopic Maps on a Curved Cortical Surface to Predict Visual Field Defects

Visual aura is a common migraine symptom. Prediction of their spatio-temporal evolution can help to understand the underlying etiology. Self-organization is used on the curved surface of the visual cortex to reproduce human retinotopy. The neural layer of the SOM algorithm is created by a novel approach to implement centroidal Voronoi tesselations on meshes. Realistic retinotopic maps are obtained by driving self-organization with training sets derived from commonly accepted magnification principles. Whereas the curved nature of the visual cortex is found to be responsible for correct map orientation.