I have a subject run through recon-all that has gross errors in GM/WM surfaces. I did my best to follow the ControlPoints tutorial on:
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/ControlPoints
My first try had >300 control points with no difference in the WM.mgz file. I stumbled upon the freesrufer.failure_modes.ppt and it states:
"Don’t edit too much! This will reduce reliability and is almost never needed. Usually this means you need to start over as you’ve done something wrong (e.g. put control points in the wrong place)."
I deleted my control points .dat file and tried again with only 12 control points, somewhat logically spaced throughout the whole brain. Still no improvement. Attached is a coronal image, typical of the error. So, what to do? Much of the normalized volume has values of 110 and it's my understanding that adding control points to values <110, that are known WM, will bring down the average value/tissue intensities. Here's a part of the stdout as I ran recon-all -autorecon2-cp -autorecon3 -subjid "subject":
preserving editing changes in output volume... doing initial intensity segmentation... using local statistics to label ambiguous voxels... computing class statistics for intensity windows... WM (104.0): 104.1 +- 6.4 [80.0 --> 125.0] GM (74.0) : 72.4 +- 9.9 [30.0 --> 96.0] setting bottom of white matter range to 82.3 setting top of gray matter range to 92.2 doing initial intensity segmentation... using local statistics to label ambiguous voxels... using local geometry to label remaining ambiguous voxels...
My control points fall withing the WM range, as does much of the "110" labeled voxels, but the surface/WM.mgz file isn't improving. What am I doing wrong and what can I try next?
Thank you in advance for any/all responses.
-Joel
