Hi Greg
you could threshold then look at Hausdorff distance of the blobs. Or you could smooth before computing correlation as that will take spatial stuff into account.
Or if you have a patch that you want to cross-correlate against the rest of the surface you could do it on the sphere, although that's a bit of a pain since you have to take the metric tensor into account
cheers Bruce
On Tue, 21 Jun 2016, GREGORY R KIRK wrote:
Hi,
has anyone created functionality to compare the spatial correlation/similarity
of two patterns in the form of cortical surface overlays. What I am looking at is a have
scalar at every point on the cortical surface. obviously this is just a list of vertices with
a scalar value associated with each. obviously I can map these both onto an average cortical surface and then simply compute the correlation of both of these.
but of course pattern similarity is an inherently spatial problem and for cortical
surfaces a 2 dimensional spatial correlation or other similarity metric would be
what you want. Here the order of vertices in a vertex list maps somewhat like a
string around the cortex and so the next wrap around you have vertices which are
adjacent on the surface but distant by the metric of vertex count.
is there any implementation of a similarity metric which takes this coordinate problem into account. In my measure every vertex across both hemi has a value and so i am not interested in something for a small localized patch.
any thoughts appreciated
Greg