Abstract
Retinotopic mapping traditionally presents simple visual stimuli (e.g. drifting bars, expanding rings), yet the technique is limited by the stimulus and the need for fixation. It is also inefficient, as voxel-by-voxel measurements of retinotopy do not leverage the strong prior of spatial organization in early visual areas. Here, we describe a method to simultaneously discover the retinotopic organization of visual areas V1-V3 without the need for fixation or traditional retinotopic mapping techniques. Three subjects underwent MRI at 7T. From the MPRAGE anatomical image of each hemisphere for each subject, a set of possible retinotopic templates for V1-V3 were generated. Each template varied in the precise location and orientation on the cortical surface, occupying a point in a three dimensional template parameter space (Benson et al. 2012; 2014). During BOLD fMRI scanning (1.5mm3, TR=2s) subjects freely-viewed 30 minutes of Pixar animated shorts (12.5° width visual angle). The time series data from each occipital vertex was used to create a cross-correlation matrix, with structure in this matrix induced by the shared visual stimulation experienced by vertices that represent similar points in visual space. Each possible retinotopic template also predicts a particular pattern of correlation across vertices. For each hemisphere, a search was conducted incorporating cortical receptive field sizes and the retinotopic templates to find the predicted correlation matrix that best fit the time series data. For all 6 hemispheres, the best fitting template was very similar to the retinotopic organization measured by traditional techniques. These results demonstrate that retinotopy can be obtained for areas V1-V3 using complex and continuous 'naturalistic' visual input, without the need for fixation, and allows for retinotopic mapping in subjects with difficultly fixating on a central target due to behavioral or ophthalmologic limitations.
Meeting abstract presented at VSS 2016