pycortex
pycortex copied to clipboard
cortical *thickness imported from freesurfer is negative
But otherwise appears right. Is this a bug in how the data file is read?
No, I think that's just how freesurfer writes the data (4 byte floating point big-endian). The curvature is also negative IIRC. I could potentially invert all of them -- what's the behavior of cortical thickness?
I think I mis-wrote this in haste, but here's the situation. Cortical thickness imported from freesurfer is negative (only negative values), which is weird and nonsensical. Sulcal depth is negative for deep sulci and positive for gyri. That seems counterintuitive given the name "depth" (we don't usually think of a pool as, e.g. -8 ft deep). Curvature is negative for sulci and positive for gyri, which seems sensible (and our curvature code returns the same polarity).
Relatedly, I tried comparing the cortical thickness estimated by freesurfer to the cortical thickness we estimate (which is just the euclidean distance from wm to pia). There are some real differences between the two that make me a little worried. Do we know how freesurfer computes cortical thickness?
In particular it seems like freesurfer truncates at 5mm, and it sets some values to zero. It also looks like our estimate is almost always an overestimate vs. freesurfer. Perhaps instead of assuming matching pia-wm vertices, they find the distance to the closest vertex? Or something fancier?
We got distracted yesterday by the thickness computation issue (and what it implied about our through-cortex sampling). I'll open that as a separate issue. But our conclusion was that we should use freesurfer's thickness estimate. And we should reverse it, so it has positive values. What about the sulcal depth? Should it be negative or positive for sulci?
No strong opinion.
On Wed, May 14, 2014 at 9:07 AM, Alex Huth [email protected] wrote:
We got distracted yesterday by the thickness computation issue (and what it implied about our through-cortex sampling). I'll open that as a separate issue. But our conclusion was that we should use freesurfer's thickness estimate. And we should reverse it, so it has positive values. What about the sulcal depth? Should it be negative or positive for sulci?
— Reply to this email directly or view it on GitHubhttps://github.com/gallantlab/pycortex/issues/82#issuecomment-43100944 .