Hi Aaron,

this is probably due to the partial volume correction we do when computing the volumes. Basically we assume that each voxel is a linear combination of the tissue at that voxel and the next most likely label that neighbors that voxel. The weighting is computed as mixture weight for the local class mean for each class that explains the voxel intensity. Our internal studies show that this improves test-retest reliability and presumably also accuracy.

cheers, Bruce

On Thu, 7 Sep 2006, Goldman, Aaron (NIH/NIMH) [C] wrote:

Hi,

I had a question regarding the total volumes of aseg labels listed in aseg.stats. We've noticed slight discrepencies between the values reported in the text file, and values obtained by converting aseg.mgz to analyze (using mri_convert) and adding up the same structures ourselves in matlab. The differences ranged from 0-2%, and varied in directionality.

To give a more specific idea of what we did, we added up all of the wm labels in the stat file to get one value, and all the gm to get another. We then created binary masks for the same sets of labels, and calculated total volume of these images in matlab.

I was wondering if the stats in aseg.stats were simple voxel counts (as I'd previously assumed since they're all integers), or if there were additional nuances to the calculation (in which case these discrepencies would presumably be expected).

Thanks,

-Aaron-

---

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

Hi Doug, Thanks for your answer. I've processed my data as you suggested, and it seems to work fine now. If I may, I would like to ask some follow-up questions to you and the list. The data I'm analysing comprises a group of healthy controls and patients with increased cortical thickness. When I compare the stats tables generated from these groups, it is obvious that there is a thickness difference in almost all of the regions studied (in eg. lh.aparc.a2005.stats). After glmfit, however, there are only a few regions of significant difference, and I suspect that this is due to the DOSS option to glmfit. If I understand correctly, then this would remove the difference of mean cortical thickness, which I am interested in here. I wonder what the best way around this is. If I test for significant difference in overall thickness seperately, can I assume that the changes displayed are those that occur over and above what the difference in mean thickness explains? More generally, I wonder if the "different offset" approach I've been running so far is correct for comparison of thickness measurements (since the difference of the mean is meaningful here). Your input is very much appreciated. Thanks a lot, Johannes

Doug Greve wrote:

Use lh.curv instead of lh.avg_curv. For lh.thickness, there is a bug in make_average_subject (resamples the avg thickness to the first or last subject instead of the average subject). If you are going to be doing a group stat analysis, you will usually get the mean thickness. Otherwise, you can just run mris_preproc with the --mean option.

doug

Johannes Klein wrote:

...

Hi all, I'm setting up a small study where I want to compare a subject to some matched controls. I've been following the Group analysis tutorial and I've run make_average_subject on my controls. Now, I'd like to check my data and view averaged curvature and thickness measurements. When I try to load the lh.avg_curv onto lh white in tksurfer, I get the following error: ERROR: number of vertices in XX/lh.avg_curv does not match surface (173612,163842)surfer: error reading curvature file The same applies to lh.avg_thickness However, there is a file called lh.curv that I can load. What am I doing wrong here? How can I view the average curvature and thickness measurements? Thanks a lot everyone, Johannes

---

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