Abstract
Purpose: Human visual field maps V1, V2 and V3 are well established. Mapping the ventral cortical surface has proven more difficult. In particular, the visual field coverage of the ventral V4 map (hV4) remains uncertain. One claim is that hV4 covers a contiguous region of ventral cortex adjacent to ventral V3 (V3v), and that hV4 responds to the entire contralateral visual hemifield (Wade et al, 2002). Others claim that hV4 is split, with the ventral portion excluding coverage of the visual field near the lower vertical meridian, and a second portion adjacent to dorsal V3 representing the remaining lower field (Hansen et al, 2007). We sought to clarify this issue using new, model-based receptive field mapping (Dumoulin & Wandell, 2008).
In analyzing data from individual subjects, we discovered a significant and general problem associated with measurements in ventral occipital cortex and particularly measurements near hV4. In most hemispheres the BOLD hV4 data are contaminated by artifacts from the transverse sinus (TS); this large artifact can be seen by inspection of the raw image data. In many cases, the TS artifact masks responses specifically in the region of cortex that distinguishes the hV4 models.
Methods: FMRI images were acquired while subjects viewed drifting checkerboards through a slowly moving bar aperture (field of view: 3° or 14° radius). A 2D-Gaussian population receptive field was fit to each voxel that best predicted the time series.
Results: All subjects showed a visual field map adjacent to V3v. The V3v/hV4 boundary was marked by an upper field polar angle reversal. Whether the map extended laterally to include the lower vertical meridian depended on the location of the TS artifact: When the TS was displaced from the lateral edge of hV4, the visual field coverage extended to the lower meridian, or nearly so, consistent with the hemifield model. In hemispheres in which the TS somewhat further from hV4, the measured field coverage was reduced, sometimes spanning as little as an upper quarterfield, consistent with the split V4 model.
Discussion: BOLD imaging of the ventral surface is limited by artifacts caused by the TS. However the alignment between the TS and the hV4 map varies across hemispheres. Subjects in which the TS effects are most remote from hV4 yield the most reliable measurements. In these subjects, hV4 represents all or most of the contralateral hemifield. The missing lower field representation in many subjects is likely due to limitations caused by this TS artifact.