Hi Christine
why don't you upload an example subject and we will see what we can do
cheers Bruce On Thu, 7 Aug 2014, Christine Smith wrote:
Hello,
I am working with some scans that have poor signal, particularly at the lateral aspects of the cortex (i.e., lateral temporal lobes). The measures of the cortical volumes for each of the lobes of the brain are reduced by 2-4 standard deviations in these healthy subjects compared to scans collected from healthy subjects at a different scanner. The measures of white matter for the lobes appears to be ok.
To show you the problem, I have attached two pictures. The first is of the surfaces as they come out of freesurfer (reference.jpg). The second is the same slice if you increase the brightness of the image to highlight that the signal is so low in some cortex (e.g., lateral temporal lobes) that it appears 'black' and absent with the default brightness/contrast settings in tkmedit. This missing cortex is visible at high brightness and low contrast (reference_lowcontrasthighbright.jpg).
We have been using several commands that have an effect on the pial line based on an earlier request for help.
recon-all -seg-ghi ## -autorecon2 -autorecon3 -s SUBJECT_ID OR recon-all -seg-wlo ## -autorecon2 -autorecon3 -s SUBJECT_ID
seg-ghi: maximum for gm at border with wm seg-wlo: minimum for wm at border with gm
We found that a ghi value of 60 captures all the grey matter that we were missing (see ghi60.jpg). However, for the superior part of the scans (i.e., above the temporal lobes), it also extends out the white matter such that sulci are missing or too shallow. This means that the grey matter is also very thin.
Combining seg-ghi and seg-wlo commands in the same call was not the solution:
recon-all -seg-ghi 60 -seg-wlo 60 -autorecon2 -autorecon3 -s SubjectID
The resultant image had the same problems with the white matter. I have examples of the scans after seg-ghi 60 alone, seg-wlo 60 alone, and both commands together, if these would be useful. We have experimented with changing the values after each these commands and had had little success identifying how they talk to each other and what the magic combination of values might be, if it exists.
If these weren't scans from patients who were very difficult to scan in the first place, I would love to just forget about them. However, these scans are precious to us and may not be able to be reacquired. Thus, I am employing heroic efforts to salvage usable data from them.
I am open to a solution that might come from two separate runs of freesurfer: one to get the white matter boundary (as it appears in the reference scan attached) and one to get the pial boundary (as it appears after using seg-ghi 60 alone); if this is a reasonable course of action.
Thank you for any support you may provide.
Best, Christine -- Christine N. Smith, Ph.D. Department of Psychiatry University of California, San Diego