yes, that is a consequence of partial volume correction. If you have a 3 class boundary (two of which are hippocampal subdivisions), we will compute different partial volume fractions than if you have a 2 class boundary. I wouldn't expect it to be a huge effect, but they definitely won't be identical
On Thu, 17 May 2012, Joshua Lee wrote: > We discovered this while subdividing the hippocampus into several subregions > using tkmedit to relabel voxels, and finding that these subregions were not > adding up to the original volume. I expected some differences, but many > deviated more than I had expected. This led me to the test I discussed at > the beginning of the thread. > > > Joshua > > > > > > > On May 16, 2012, at 1:55 PM, Bruce Fischl <fis...@nmr.mgh.harvard.edu> wrote: > >> hmmm, that is convincing. I don't think there's anything in mri_segstats >> that knows about the meaning of any of the indices, but maybe Doug can >> correct me? >> >> On Wed, 16 May 2012, Joshua Lee wrote: >> >>> Hi Bruce, >>> First, let me add that if I run the mri_segstats without the --pv flag, the >>> two will have identical volumes. This indicates to me that we captured all >>> the voxels into the new label. >>> Joshua >>> - >>> Joshua Lee >>> Graduate Student >>> Center for Mind and Brain & >>> Department of Psychology >>> University of California, Davis >>> 530.747.3805 >>> On Wed, May 16, 2012 at 1:38 PM, Bruce Fischl <fis...@nmr.mgh.harvard.edu> >>> wrote: >>> Hi Joshua >>> >>> probabilities are computed internally, but not stored nor used >>> in volume calcualtions. Are you actually doing a 3D fill? I >>> don't think the hippocampus label is guaranteed to be connected. >>> Can you try instead replacing every voxel in matlab with the >>> hippocampus index to your new one? That I think should give you >>> the same results. >>> >>> cheers >>> Bruce >>> >>> On Wed, 16 May 2012, Joshua Lee wrote: >>> >>> Suppose I were to open a subject aseg.mgz in >>> tkmedit, and use the 3D fill >>> tool to relabel the entire left hippocampus >>> segmentation to >>> 'MyNewImaginaryLabel'. If I then re-run the aseg >>> stats, should the left >>> hippocampus volume from the original aseg.stats be >>> identical to the volume >>> of MyImaginaryLabel in the newly generated >>> aseg.stats? >>> >>> Every voxel from the prior label is now assigned to >>> the new label; yet, I >>> am getting different results (by about 100 voxels in >>> the ones I've tried). >>> >>> The command Im using to recalculate: >>> mri_segstats --seg >>> /usr/local/freesurfer/subjects/s02/new_MRI/new/aseg.auto.mgz >>> --sum >>> >>> /usr/local/freesurfer/subjects/s02/stats/NEWmethod1aseg_Hipp.stats >>> --pv >>> /usr/local/freesurfer/subjects/s02/mri/norm.mgz >>> --excludeid 0 >>> --brain-vol-from-seg --brainmask >>> /usr/local/freesurfer/subjects/s02/mri/brainmask.mgz >>> --in >>> /usr/local/freesurfer/subjects/s02/mri/norm.mgz >>> --in-intensity-name norm >>> --in-intensity-units MR --etiv --surf-wm-vol --ctab >>> /usr/local/freesurfer/8regions.txt --subject s02; >>> >>> Now if I were to relabel 'MyNewImaginaryLabel' to >>> 'AthirdLabel', and again >>> changing every voxel in 'MyNewImaginaryLabel' to the >>> new label, the computed >>> volumes do not change. This led me to suspect that >>> aseg.mgz contains >>> probability estimates at each voxel that are then >>> subsequently overwritten >>> by my tkmedits. However, per a previous conversation >>> from a year ago or so >>> with Bruce and also the relevant Freesurfer >>> documentation this does not seem >>> to be the case. Partial volume information is not >>> contained within the >>> aseg.mgz (right!?), but is computed by mri_segstats >>> by comparing border >>> voxels with intensities in the norm.mgz. So I am at >>> a loss on why these >>> should be different. >>> >>> This has been replicated in version 4.5 and 5.1 >>> - >>> Joshua Lee >>> Graduate Student >>> Center for Mind and Brain & >>> Department of Psychology >>> University of California, Davis >>> 530.747.3805 >>> The information in this e-mail is intended only for the person to whom >>> it is >>> addressed. If you believe this e-mail was sent to you in error and the >>> e-mail >>> contains patient information, please contact the Partners Compliance >>> HelpLine at >>> http://www.partners.org/complianceline . If the e-mail was sent to you >>> in error >>> but does not contain patient information, please contact the sender >>> and properly >>> dispose of the e-mail. > > > > _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
