Purchase this article with an account.
Paul VanValkenburgh; A data acquisition engineer's attempts to improve the precision of his retinotopic map. Journal of Vision 2005;5(12):91. doi: 10.1167/5.12.91.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Accurate viewing of 3-D fMRI is compromised in many ways, with the existing tools for segmentation, inflation, and flattening, particularly regarding hidden and/or distorted surfaces. Improved stability of the subject was attempted with beta-blockers, by fixation on a single target pixel, and saccade suppression. Stimulus distortion can be reduced by more precise control of the stimulus and mirror distances and angles. Manual analysis of activated voxel bands was simplified by culling all non-activated and smoothed structural data, and using realtime rotation and zoom viewing of the 3-D “point-cloud” data. It also helped to follow the ring stimulus bands, which appear to be continuous across many visual cortex areas, as opposed to intermittent ray bands. Nodes for a wireframe model, were defined by manually selecting 3-D co-ordinates of the intersection (dot-in-box) of the ring-or-ray stimulus activated voxels. The cortex surface is approximated by quadrilateral facets, with their sides defined by true distances from node to node. Realtime “flythrough” semi-transparent 3-D plots in Matlab were used to view the backside of hidden cortex surfaces. Finally, a large-scale physical model of V1 was constructed, to resolve mathematically indeterminate unwrinkling of cortex folded in the calcarine sulcus.
This PDF is available to Subscribers Only