Abstract
Retinotopic maps estimated via population receptive field (pRF) mapping (Dumoulin et. al., 2008) are influenced by various factors, such as visual stimulus design, fMRI acquisition parameters, preprocessing, analysis and finally, thresholding. Usually decided on a best-practice basis, a threshold of explained variance should remove unreliable voxels, but still retain comprehensive pRF coverage of the visual field in healthy subjects. To this end, we use “quiver plots” to visualize the stability of estimated pRFs between runs based on different explained variance thresholds. Six healthy subjects (age: 25.8 ± 4.5 years; 3 female) were measured on a 3T Siemens TIM Trio scanner. The CMRR multiband sequence (Moeller et. al. 2010) was used to acquire two functional runs with TE/TR = 36ms/1500ms, voxel size = 1mm3, 28 slices, MB = 2 and 224 volumes per run. The stimulus consisted of a flickering, traveling bar covering the central 18.8° visual angle. A structural MPRAGE image was recorded and segmented with the help of Freesurfer. Functional data was slice-time corrected and realigned using SPM12. PRF maps were estimated on gray matter voxels with mrVista. We created quiver plots by connecting estimated run 1 and 2 pRF centers of the same suprathreshold V1 voxels, where arrows indicate the direction of change. For a typical subject, increasing the threshold of explained variance leads to smaller average pRF center distances (0.98° ± 1.61° visual angle for a threshold of 0%, 0.54° ± 0.41° for 10% and 0.40° ± 0.26° for 50%). As pRF stability increases, pRF coverage of the visual field is reduced. Quiver plots can not only help to determine a sensible threshold of explained variance, but also to visualize differences of pRF mapping runs differing by stimulus, fMRI sequence, preprocessing, analysis or involving artificial scotomata.