Thanks, Doug!
So I tried this method using fsaverage/surf/lh.white.avg.area.mgh. The results seem to be reasonably consistent with the ones I get when eyeballing the p-maps overlayed on fsaverage in tksurfer. Would you agree that's a reasonable "sanity check"?
When I compare the results to just counting vertices, there doesn't seem to be a huge difference between proportion of #vertices and proportion of surface area. So either there isn't a lot of variance in the sizes of the faces, or the variance is evenly distributed.
Finally, if I want to convert from an individual map or the cohort mean map to mm² units, can I do it by scaling the ico area vertex values by the ratio between native and ico total area? So: (vertex value from mean cohort map * total area for cohort mean map) / total area for fsaverage/surf/lh.white.avg.area.mgh, where total area would be the sum of all vertex values?
Thank you!
LMR
-------------------------------------------------------------------------------------------------------
You don't want to get the surface area from the faces of fsaverage. Instead use the values in fsaverage/surf/lh.white.avg.area.mgh (this is the average of the group used to create fsaverage), or, probably better, get an average area map for your cohort doug
On 07/30/2014 08:38 AM, Lars M. Rimol wrote:
Hi Bruce,
I would like to be able to tell what proportion of a region of interest (ROI), as defined in atlas space by e.g. Desikan-Killiany, that shows a significant effect (based on a p-map). For now I overlay the p-map on the inflated surface of fsaverage in tksurfer and eyeball the proportion.
Given a p-map, if I find the FDR threshold and identify the vertices within a given ROI that have a p-value greater than the threshold, then I can find the proportion of the ROI that is suprathreshold. E.g., I find 1986 suprathreshold vertices in "bankssts" out of 2137, so 93% of vertices in bankssts show a significant effect.
My question is: Does that tell me anything about what proportion of the ROI's surface area is affected in atlas space? Obviously, if the faces were uniform, there would be a 1 to 1 relationship between #vertices and area. In the original tesselation of any dataset the size of the faces is uniform, but that changes with topology fix and deformation. I assume that is true also for fsaverage? (so I can't assume [#sig vertices] / [# tot vertices] == the proportion of the ROI's area that is significant in atlas space)
I can find the surface area of the suprathreshold region for my sample (or any subset thereof) by looking at a mean area map. But I'm unsure how to do that for fsaverage itself? Is there information on regional surface area directly available? Or would using getFaceArea.m or getFacesArea.m or similar functions be a solution?
Thank you!
On 8/8/14 10:40 AM, Lars M. Rimol wrote:
Thanks, Doug!
So I tried this method using fsaverage/surf/lh.white.avg.area.mgh. The results seem to be reasonably consistent with the ones I get when eyeballing the p-maps overlayed on fsaverage in tksurfer. Would you agree that's a reasonable "sanity check"?
I'm not sure what you mean here
When I compare the results to just counting vertices, there doesn't seem to be a huge difference between proportion of #vertices and proportion of surface area. So either there isn't a lot of variance in the sizes of the faces, or the variance is evenly distributed.
The vertex sizes may be fairly consistent because each vertex area is the average over 40 subjects.
Finally, if I want to convert from an individual map or the cohort mean map to mm² units, can I do it by scaling the ico area vertex values by the ratio between native and ico total area? So: (vertex value from mean cohort map * total area for cohort mean map) / total area for fsaverage/surf/lh.white.avg.area.mgh, where total area would be the sum of all vertex values?
Why not just compute the sum of the cohort mean area map overt the vertices of interest? doug
Thank you!
LMR
You don't want to get the surface area from the faces of fsaverage. Instead use the values in fsaverage/surf/lh.white.avg.area.mgh (this is the average of the group used to create fsaverage), or, probably better, get an average area map for your cohort doug On 07/30/2014 08:38 AM, Lars M. Rimol wrote:
Hi Bruce,
I would like to be able to tell what proportion of a region of interest (ROI), as defined in atlas space by e.g. Desikan-Killiany, that shows a significant effect (based on a p-map). For now I overlay the p-map on the inflated surface of fsaverage in tksurfer and eyeball the proportion.
Given a p-map, if I find the FDR threshold and identify the vertices within a given ROI that have a p-value greater than the threshold, then I can find the proportion of the ROI that is suprathreshold. E.g., I find 1986 suprathreshold vertices in "bankssts" out of 2137, so 93% of vertices in bankssts show a significant effect.
My question is: Does that tell me anything about what proportion of the ROI's surface area is affected in atlas space? Obviously, if the faces were uniform, there would be a 1 to 1 relationship between #vertices and area. In the original tesselation of any dataset the size of the faces is uniform, but that changes with topology fix and deformation. I assume that is true also for fsaverage? (so I can't assume [#sig vertices] / [# tot vertices] == the proportion of the ROI's area that is significant in atlas space)
I can find the surface area of the suprathreshold region for my sample (or any subset thereof) by looking at a mean area map. But I'm unsure how to do that for fsaverage itself? Is there information on regional surface area directly available? Or would using getFaceArea.m or getFacesArea.m or similar functions be a solution?
Thank you!
-- yours,
Lars M. Rimol, PhD St. Olavs Hospital Trondheim, Norway
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas Greve -
Lars M. Rimol