Hello all Freesurfer users!
I am planning to do some cortical thickness analysis in children affected by psychiatric disorders, and after a few months of reflection, I still have some trouble to obtain a correct segmentation in the parietal region with my images.
Actually, I am especially interested in the parietal lobe where we already observed a decrease in the gray matter volume in this sample of patients, but I have some trouble with the segmentation of the gray-white matter in this region. In fact, my orig and smoothwm surfaces look pretty good, except in about 10 slices in the parietal lobe where parietal cortex is so bright that it is classified as white matter. I first thought that it was either due to the quality of the scan (1.5 Tesla, slices 1.5mm) or related to a bias in the magnetic field. So first I tried to apply different bias correctors, but it has no effect on the segmentation. Finally, I found that forcing the high gray and white low values during the segmentation (in the Wm Filter Expert Preferences) was really interesting to obtain a better orig surface on my images. What I did is I applied a Gaussian Fitting algorithm of segmentation on the normalized and skull stripped image (brain), and took the limit values of the partial volume intensity for the high gray and low white value, so that the segmentation included the tissue classified as partial volume (about 97 for white low and 104 for high gray, depending on the subjects). This had a quite benefic effect on the parietal cortex, but as you can probably imagine, it excluded the subcortical structure of my wm images, and increase drastically the time spent for manual editing. Further, most of the time I have to delineate the insular cortex manually, so I just displaced the segmentation problem from the upper to the medial structures.
I am quite perturbed by this failure to obtain a correct cortical ribbon in the parietal region with my images, any advice or idea will be greatly appreciated…
Are my images not enough good for what I want to do? Did someone already notice that parietal cortical thickness was sometimes difficult to assess? (I checked the bert’s one, which also show a really thin parietal cortex?)
How can I solve this problem? Is there any way to apply differentially the rules for the segmentation according to the region in Freesurfer? For example, based on a reference atlas that specifies that partial volume have to be included in the parietal region, and excluded in the subcortical region? Or referring to the talairach boxes, that means specify the high gray and low white value for all the regions that correspond to the coordinates of the parietal talairach boxes?
Thank you very much for your help,
Best regards,
Marie Schaer Neuroimaging Laboratory Child Psychiatry Department Geneva Faculty of Medicine Switzerland
PS: I attached the images with my first message, but it was refused because too big for the mailing list, but I can easily provide you the images again.