Abstract
FMRI studies of the visual system often rely on retinotopy to define the low-level visual areas (V1, V2, V3, hV4) and to establish a correspondence between voxels in these areas and points in the visual space. However, retinotopic mapping is impossible for blind subjects, subjects who cannot fixate, or those with severely compromised visual fields. Here, we evaluate a method for inferring the retinotopy from normally-sighted subjects using a landmark-based cortical surface coregistration technique (Pantazis et al., 2010). This technique relies on manually identified sulcal fundi to constrain the coregistration progress.
Sagittal T1-weighted images of 1-mm isotropic spatial resolution were obtained from five normally-sighted individuals and used to derive a surface representation of their cortices. Retinotopies were obtained using standard methods. Seven sulci in or close to the occipital lobe were manually labelled for inter-subject coregistration. For each subject, we generated a “foreign” retinotopy by coregistrating the subject against the four other subjects. The eccentricity and polar-angle maps of the four subjects were projected to the target subject and averaged. Based on these averaged foreign maps, the boundaries between visual areas were drawn using the standard approach. This process was repeated for all subjects.
Across the subjects, we found that the mean absolute deviation between the natively and foreignly defined visual-area boundaries was 3.7 ± 1.9 mm on a Talairach-normalized cortex. The mean distance between the centers of the foveal confluence was 7.8 ± 5.6 mm. Expressed in cortical distance, the native eccentricity was 1.31 ± 0.3 times the foreign eccentricity, and this scaling factor was marginally consistent across the cortex (r = 0.74). These results suggest that we can reasonably infer retinotopy and identify the low-level visual areas based on the landmark cortical coregistration, when a native retinotopy is unavailable.
National Institutes of Health Grant R01-EY017707.