Subject: Re: [Freesurfer] Cortical and White Matter Parcellation and Segmentation Outputs From: lbotelho@nmr.mgh.harvard.edu Date: Tue, February 12, 2008 4:09 pm To: "Bruce Fischl" fischl@nmr.mgh.harvard.edu --------------------------------------------------------------------------

Hi,

1. Some samples with the cortex/gray matter gaps in the parcellation results. Is the mri_convert --fill_parcellation utility useful here?

2. My interest and use is focused on aseg+aparc and wmparc too, thanks for pointing this clearly.

3. The sample in attachment is less than optimal (though a 3T MPRAGE, 1mm3) but to illustrate the challenge in this kind of data. Thanks for the suggestion of control points. Can you point me some reference about its usage?

Best, Luis

Hi Luis,

1. I'll leave this for Doug.
2. The aseg.mgz is indendent of (and produced before) the mri_segment

step (which produces the wm.mgz). In general we don't use the cortical part of the aseg as the surfaces in our experience are more accurate. 3. Can you give an example of this? I would think control points could help. Note that we don't use image intensity, but rather local gradient maximum to position the pial surface, which should be independent of local bias.

cheers, Bruce

On Mon, 11 Feb 2008 lbotelho@nmr.mgh.harvard.edu wrote:

...

Dear FS experts,

1. There are some frequent mismatches between the white matter and

cortex parcellations (aparc+aseg.mgz and wmparc.mgz) with zeroed voxels between them.

Would suppressing the option --ribbon when running aparc2aseg either during cortex or white matter parcellation (or both) result in a more close match between aparc+aseg.mgz and wmparc.mgz? Other suggestions?

2. Cortex segmentation output in aseg.mgz is not uncommonly thinner than

it should be in my data. Tweaking around the default mri_segment settings for white matter lo/hi and gray matter hi may improve it. I am still looking for optimal settings and the pial surface is not under control. Any specific suggestions?

3. These days acquiring with multichannel headcoils has become almost

the norm. Even normalized data seems to suffer somehow. Particularly, superficial and brighter cortex is prone to produce thinner cortex thickness results than cortex in basal regions with less signal. Any workarounds?

Many thanks, Luis

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer