Effects of image coherence on the activity of posterior visual areas: an fMRI study

Natural images have a characteristic 1/f a power spectrum (Field, 1985). Different natural images thus differ mostly in their phase, and their spatial structure (e.g. edges) results in correlated phase spectra. Our aim was to study how processing in visual cortex varies as a function of phase correlation. To do this, we generated random phase spectra and performed linear interpolation between these random spectra and phase spectra of natural images. By using different interpolation levels (0,25,50,75,100 signal), we could parametrically investigate the effects of phase correlation on the activity in visual areas, while keeping the power spectrum constant. Images were 10°×10° in size, and were presented at randomized locations within ±3° of the center of gaze to mimick saccadic eye movements. To study such effects by means of BOLD imaging, multi-shot (8) horizontal GE-recalled EPI images (voxel volume 0.5×0.5×2mm3, TE=20, TR=740, FA=50) of the entire brain of anesthetized monkeys were obtained using a Biospec 4.7T/40cm scanner (Bruker, Inc.) with 50mT/m gradients. This preparation affords the unique opportunity to examine activity in many brain areas simultaneously independent of cognitive factors, and provides a baseline to which human and awake monkey fMRI data on object processing may be compared. We found that many voxels were activated by natural images and/or interpolated patterns. Activity in striate and extrastriate cortical areas was modulated by interpolation level, with natural objects at 100 signal typically eliciting most activity. We plan to examine the effects of varying power spectra (i.e. values of a) on the BOLD signal