Abstract
We compared two types of visual localizer to investigate the spatial accuracy of gradient echo BOLD fMRI at 3T. The “single-condition” localizer presented two annuli of flickering checkerboard (the target stimulus) against a mean gray background in block alternation (target for 8s, blank for 8s). The “differential” localizer comprised block alternation between the target stimulus and a checkerboard filling the complement of the visual field (i.e. everywhere except the target annuli). Each localizer was used to select a region of interest (ROI) at each of two spatial resolutions. Because a differential protocol minimizes blurring attributable to large pial and intracortical veins, we expected the differential localizer to identify a subset of the voxels selected by the single-condition localizer. However, at low-resolution (3mm isotropic voxels), a surprisingly large number of voxels selected by the differential localizer were not selected by the single condition localizer. In fact, many voxels that responded positively to the stimulus in the differential protocol responded negatively to the stimulus in the single-condition protocol. Estimating the hemodynamic impulse response function (HIRF) and mean-normalized variance of individual voxels demonstrated that the localization errors could be attributed to voxels near large veins. A much smaller percentage of the voxels in ROIs defined at high-resolution (1.2mm) exhibited these errors. These results reveal a distinct advantage to fMRI measurements at high spatial resolution, in spite of the lower signal-to-noise ratio, above and beyond the expected improvement in spatial accuracy due to finer spatial sampling.
McKnight Foundation, Seaver Foundation