[pymvpa] searchlight radius vs. mask

[Michael Hanke]

On Fri, Mar 05, 2010 at 04:35:58PM -0500, kimberg at mail.med.upenn.edu
wrote:
> I have a quick question about searchlights.  I'd like to do an analyis
> where the sphere centers are restricted by one mask and the sphere
> extents are restricted by another mask.  More specifically, I want to
> do the searchlight for each voxel in a slice, but when it's doing each
> voxel, the searchlight radius should not be restricted to that slice,
> just restricted to the brain.  Is this an off-the-shelf capability?

Yes it is.

http://dev.pymvpa.org/examples/searchlight.html

Quote:
| In this example, we do not want to compute full-brain accuracy maps, but
| instead limit ourselves to a specific subset of voxels. We’ll select all
| voxel that have a non-zero z-stats value in the localizer mask we loaded
| above, as center coordinates for a searchlight sphere. These spheres
| will still include voxels that did not pass the threshold. the localizer
| merely define the location of all to be processed spheres.

Just come up with a slice mask to replace the z-stats.

HTH,

michael

-- 
GPG key:  1024D/3144BE0F Michael Hanke
http://mih.voxindeserto.de