Abstract
Ongoing debate focuses on the functional properties of human object- and face-selective visual cortex. Recent reports disagree about whether high-resolution (approximately 1mm3 voxels) and standard-resolution (approximately 27mm3 voxels) fMRI reveal qualitatively different patterns of activation in these regions. Here, we investigated face- and object-selective fMRI activations at high-resolution to establish their reproducibility and dependence on a variety of scanning parameters. Five subjects participated in four scanning sessions in which we acquired fMRI data at high resolution (1.5mm isotropic voxels, 3 Tesla, bilateral prescription). Each subject participated in four sessions over several weeks using each of two radio-frequency coils: a single-channel surface coil, or an 8-channel surface coil array and each of two T2*-weighted acquisitions: either slice or slab excitation. During scanning, subjects viewed six image types: grayscale front-view faces; grayscale face profiles; grayscale abstract sculptures; two-tone face silhouettes; two-tone abstract shapes; and scrambled images. We examined the reliability of object- and face-selective activations across sessions within large anatomically defined regions including the posterior fusiform gyrus and lateral occipital cortex. Our results reveal robust, reproducible patterns of response within subjects across sessions for contrasts of object vs. scrambled images in lateral occipital cortex (R= 0.22±0.11, p−3) and face vs. object images in fusiform (R= 0.47±0.06, p−8). We also found reliable profiles of the mean time course (R= 0.74±0.17, p−13) and its stimulus selectivity (R= 0.81±0.18, p−13) across functionally defined regions of interest including object- and face-selective patches in lateral occipital cortex and fusiform gyrus. Additional analyses revealed differences in the volume of activation associated with different scanning parameters. However, these differences varied by the region and contrast of interest. In summary, our results demonstrate that standard methods for determining object- and face-selective activations are robust and reproducible across sessions using high-resolution fMRI.
This research was funded by the following grants: Whitehall 2005-05-111-RES NEI 5 R21 EY016199-02 and by NIH P41 RR09784.