Abstract
Most methods for assessing the spatial selectivity of visual areas using fMRI employ periodic stimuli and rely on well-preserved retinotopy. However, it is possible that a population code across a region might reliably represent the spatial position of a stimulus, even in the absence of strong retinotopic organization. Here, we used a forward encoding model which provides an estimate of the BOLD response of hypothesized information channels, modeled here as 2D spatial receptive fields, to evaluate spatial coding across occipital, parietal, and frontal regions. Observers viewed flickering checkerboards presented at different locations on the screen. We then reconstructed the visual stimulus viewed by each observer using the BOLD signal in each independently localized area. We could produce stable and robust reconstructions from responses in occipital visual areas, posterior IPS, and human MT (hMT+). These reconstructions are shown to be reproducible across sessions and are generalizable to untrained stimuli, and regions known to have larger receptive fields are shown to have coarser visual field reconstructions. The fidelity of reconstructions achieved with this model, which uses significantly fewer information channels than many others (Nishimoto et al, 2011), allows for swift data acquisition and supports the measurement of spatial maps across different task demands using identical stimuli. This technique can be used to probe models of information coding across cortical regions and to ask novel questions about the dynamic neural mechanisms of cognitive processes, like visual attention.
Meeting abstract presented at VSS 2012