segmentation vs parcellation: Segmentation appears to mean identifying and assigning voxels to a label which maps to a gross structure like thalamus, left cortex, etc. Parcellation appears to mean the same but for mappings at higher resolution, e.g. L V1. Is this correct?
surface vs volume: My questions here arise from differences in the results of extracting labels when using mri_cor2label vs mri_annotation2label. Spot checks of mri_cor2label results show sets of pixels on a 1 mm grid which fill the volume associated with the label. I have been assuming that this is "volume" although volume also refers to what is captured in the .mgz files.
Spot checks of mri_annotation2label results show sets of pixels which appear to delineate boundaries or "surfaces" of the gray matter (--surface orig), white matter ( --surface white) or pia (--surface pial). There appears to a similar switch in the help text for mri_cor2label, i.e. -surf subject hemi <surf>, but it appears to specify the nature of the input rather than that of the output as does the -surface switch in mri_annotation2label. This leads to my next question regarding the input "volume" but first, is what I've said correct so far? An addition concern that I have is that "surface" may be used in reference to a projection onto the sphere.
input volumes: Here I have only tried the .mgz files contained in directory ${FREESURFER_HOME}/subjects/NNN/mri . Which of these files are useful as input to mri_cor2label for outputting .label files that are likely to accurately fill the structures to which the label corresponds? What I mean here is which ones represent the end-points of a processing stream and which ones are intermediates? I understand that the intermediates are likely useful given that they are being saved in the standard processing streams.
If I load them into freeview, specify the standard color map, and place the cursor on the cerebellum, I can see which ones give the right answer, e.g. aparc.a2009, wmparc, aparc+aseg, aseg.auto, and which don't, e.g. filled, wm.seg, wm, rh.ribbin, brain, brain.finalsurfs, T1, orig (of course). But "filled", for instance, does appear to provide a good segementation of the left and right supra-tentorial white matter although the colors (right=127; left=255) do not map to those names in the standard color table.
Regards,
Don
Don Krieger, Ph.D. Department of Neurological Surgery University of Pittsburgh (412)648-9654 Office (412)521-4431 Cell/Text
On 08/12/2014 01:25 PM, Krieger, Donald N. wrote:
_segmentation vs parcellation_:
Segmentation appears to mean identifying and assigning voxels to a label which maps to a gross structure like thalamus, left cortex, etc. Parcellation appears to mean the same but for mappings at higher resolution, e.g. L V1. Is this correct?
Actually, they basically mean the same thing. "segmentation" emerged from our volume tools and parcellation emerged from our surfaces tools. The segmentation format is a single index assigned to each voxel that then indexes a look up table where it gets color and name. The parcellation format stores an RGB at each vertex, then uses this RGB to index into a LUT to get the name.
_surface vs volume_:
My questions here arise from differences in the results of extracting labels when using mri_cor2label vs mri_annotation2label.
Spot checks of mri_cor2label results show sets of pixels on a 1 mm grid which fill the volume associated with the label.
I have been assuming that this is “volume” although volume also refers to what is captured in the .mgz files.
Right, when a label is derived from a volume, the xyz must be on the grid of the volume. For surfaces, they must be on the surface mesh, but this is not a grid. You'll also notice that labels derived from volumes will have -1 as the first column (the vertex index, which makes no sense for volumes, of course, so -1)
Spot checks of mri_annotation2label results show sets of pixels which appear to delineate boundaries or “surfaces” of the gray matter (--surface orig), white matter ( --surface white) or pia (--surface pial).
There appears to a similar switch in the help text for mri_cor2label, i.e. –surf subject hemi <surf>, but it appears to specify the nature of the input rather than that of the output as does the –surface switch in mri_annotation2label. This leads to my next question regarding the input “volume” but first, is what I’ve said correct so far? An addition concern that I have is that “surface” may be used in reference to a projection onto the sphere.
This gets back to your first question about segmentation vs parcellation. In FS, parcellations are always on the surface. However, segementations can be volume or surface. If you have a segmentation on the surface and want to create a label, then you use cor2label and spec the --surf. This will fill the first column of the label with the vertex number instead of -1 and the xyz will derive from the surface you give it.
_input volumes_:
Here I have only tried the .mgz files contained in directory ${FREESURFER_HOME}/subjects/NNN/mri .
Which of these files are useful as input to mri_cor2label for outputting .label files that are likely to accurately fill the structures to which the label corresponds?
What I mean here is which ones represent the end-points of a processing stream and which ones are intermediates?
I understand that the intermediates are likely useful given that they are being saved in the standard processing streams.
aseg.mgz for non-cortical structures. For cortical structures, you can use ?h.aparc.annot (or possibly aparc+aseg.mgz)
If I load them into freeview, specify the standard color map, and place the cursor on the cerebellum, I can see which ones give the right answer, e.g. aparc.a2009, wmparc, aparc+aseg, aseg.auto, and which don’t, e.g. filled, wm.seg, wm, rh.ribbin, brain, brain.finalsurfs, T1, orig (of course). But “filled”, for instance, does appear to provide a good segementation of the left and right supra-tentorial white matter although the colors (right=127; left=255) do not map to those names in the standard color table.
filled, brain, brainmask, brain.mask.finalsurfs are intermediate volumes doug
Regards,
Don
Don Krieger, Ph.D.
Department of Neurological Surgery
Universityof Pittsburgh
(412)648-9654 Office
(412)521-4431 Cell/Text
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Thanks very much, Douglas. That is very helpful -- definitely newbie questions ...
Regards, Don Don Krieger, Ph.D. Department of Neurological Surgery University of Pittsburgh (412)648-9654 Office (412)521-4431 Cell/Text
-----Original Message----- From: freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer- bounces@nmr.mgh.harvard.edu] On Behalf Of Douglas N Greve Sent: Friday, August 15, 2014 12:22 PM To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] language questions+
On 08/12/2014 01:25 PM, Krieger, Donald N. wrote:
_segmentation vs parcellation_:
Segmentation appears to mean identifying and assigning voxels to a label which maps to a gross structure like thalamus, left cortex, etc. Parcellation appears to mean the same but for mappings at higher resolution, e.g. L V1. Is this correct?
Actually, they basically mean the same thing. "segmentation" emerged from our volume tools and parcellation emerged from our surfaces tools. The segmentation format is a single index assigned to each voxel that then indexes a look up table where it gets color and name. The parcellation format stores an RGB at each vertex, then uses this RGB to index into a LUT to get the name.
_surface vs volume_:
My questions here arise from differences in the results of extracting labels when using mri_cor2label vs mri_annotation2label.
Spot checks of mri_cor2label results show sets of pixels on a 1 mm grid which fill the volume associated with the label.
I have been assuming that this is "volume" although volume also refers to what is captured in the .mgz files.
Right, when a label is derived from a volume, the xyz must be on the grid of the volume. For surfaces, they must be on the surface mesh, but this is not a grid. You'll also notice that labels derived from volumes will have -1 as the first column (the vertex index, which makes no sense for volumes, of course, so -1)
Spot checks of mri_annotation2label results show sets of pixels which appear to delineate boundaries or "surfaces" of the gray matter (--surface orig), white matter ( --surface white) or pia (--surface pial).
There appears to a similar switch in the help text for mri_cor2label, i.e. -surf subject hemi <surf>, but it appears to specify the nature of the input rather than that of the output as does the -surface switch in mri_annotation2label. This leads to my next question regarding the input "volume" but first, is what I've said correct so far? An addition concern that I have is that "surface" may be used in reference to a projection onto the sphere.
This gets back to your first question about segmentation vs parcellation. In FS, parcellations are always on the surface. However, segementations can be volume or surface. If you have a segmentation on the surface and want to create a label, then you use cor2label and spec the --surf. This will fill the first column of the label with the vertex number instead of -1 and the xyz will derive from the surface you give it.
_input volumes_:
Here I have only tried the .mgz files contained in directory ${FREESURFER_HOME}/subjects/NNN/mri .
Which of these files are useful as input to mri_cor2label for outputting .label files that are likely to accurately fill the structures to which the label corresponds?
What I mean here is which ones represent the end-points of a processing stream and which ones are intermediates?
I understand that the intermediates are likely useful given that they are being saved in the standard processing streams.
aseg.mgz for non-cortical structures. For cortical structures, you can use ?h.aparc.annot (or possibly aparc+aseg.mgz)
If I load them into freeview, specify the standard color map, and place the cursor on the cerebellum, I can see which ones give the right answer, e.g. aparc.a2009, wmparc, aparc+aseg, aseg.auto, and which don't, e.g. filled, wm.seg, wm, rh.ribbin, brain, brain.finalsurfs, T1, orig (of course). But "filled", for instance, does appear to provide a good segementation of the left and right supra-tentorial white matter although the colors (right=127; left=255) do not map to those names in the standard color table.
filled, brain, brainmask, brain.mask.finalsurfs are intermediate volumes doug
Regards,
Don
Don Krieger, Ph.D.
Department of Neurological Surgery
Universityof Pittsburgh
(412)648-9654 Office
(412)521-4431 Cell/Text
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
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@nmr.mgh.harvard.edu
-
Douglas N Greve -
Krieger, Donald N.