FWIW, my final lists of regions that I think can be returned by default recon-all for the '2000' and '2009' parcellations are attached.
Lists come from the FreeSurfer LUT, with an extra TISS__TYPE column of (grep-able) tissue labels compiled from mri_binarize.c and also just by checking names. (The MAP_NUM column can be ignored.) The defined tissue types/labels are:
tiss__gm: gray matter tiss__wmat: white matter tiss__csf: CSF tiss__vent: ventricle + choroid plexus tiss__othr: optic chiasm, non-WM hypoint, etc. tiss__unkn: FS-defined 'unknown'
As noted before, I think the current 'mri_binarize --gm' might be updatable by some of this information; right now, the 'tiss__othr' regions seem to be included as GM.
Any feedback/comments on this list and proposed classifications would be very welcome. This is probably what we will start using in AFNI.
--pt
On Thu, Dec 15, 2016 at 8:43 PM, P Taylor neon.taylor@gmail.com wrote:
Hi-
Just gently re-bumping this question/thread.
Noting that the previous thread with recommendations about non-GM ROIs, which I used in the coming up with the GM list above, is here: http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg50005.html
Thanks, pt
On Mon, Dec 12, 2016 at 10:43 AM, P Taylor neon.taylor@gmail.com wrote:
And, a minor clarification on one of the ROIs in question--
I should have phrase the comment about ROI #80 as follows: 80 non-WM-hypointensities -------> *should be in list of FS_nongm?
... since the name clearly implies that it is non-WM.
Basically, I am first wanting to check that the above list of ROIs does form the "whole brain" returned by recon-all; there are a lot more ROIs listed in the LUT, of all sorts of tissue types, but it seems like those don't often appear in the output.
And following on that, it seems like presently a map of various non-GM tissues is made by summing up ROIs of a given type, and then: GM_map = whole_brain - WM_map - CSF_map - vent_map. However, am I interpreting things correctly that the four ROIs flagged above (#30, 62, 80 and 85) seem to be an "other" class that should be subtracted away from the whole_brain, as well, such that: GM_map = whole_brain - WM_map - CSF_map - vent_map - other_four_regions ?
Thanks, pt
On Mon, Dec 12, 2016 at 1:04 AM, P Taylor neon.taylor@gmail.com wrote:
Hi, Doug-
I've looked more into my earlier question of identifying FS-returned regions, in particular:
- which regions of the whole FreeSurferColorLUT.txt seem to get
returned by default recon-all calls (for either the 2000- or 2009-parc+segs); and 2) which of those regions are GM and non-GM (WM/CSF/other), based both on the mri_binarize list you had posted earlier in this thread and on the names given in the LUT.
I'd like to pick your brain on this.
A) The main questions I have relate to the regions labeled from [0, 256], which seem have the same labelling in both the 2000 and 2009 parc+segs. From a few different FreeSurfer recon-alls on healthy adults, below is a list of the ROIs that seem to be consistently returned by recon-all (a much sparser subset than all 256). For each, I have included the number and label, as well as put arrows by the numbers that, according to the MatchValues list in mri_binarize, should be in either the WM or Vent masks, (and hence non-arrowed regions should be GM, by how that function returns "--gm"). My main questions arise about #30, 62, 80 and 85, as they don't appear to be GM-- therefore, am I correct that I should also *exclude* those four regions from a list of GM regions? (Here, "FS_nongm*" means that it was in the list of non-GM regions in mri_binarize, as either unknown, WM, CSF or ventricle.)
0 Unknown -> FS_nongm
2 Left-Cerebral-White-Matter -> FS_nongm_wm
4 Left-Lateral-Ventricle -> FS_nongm_vent 5 Left-Inf-Lat-Vent -> FS_nongm_vent
7 Left-Cerebellum-White-Matter -> FS_nongm_wm 8 Left-Cerebellum-Cortex
10 Left-Thalamus-Proper 11 Left-Caudate 12 Left-Putamen 13 Left-Pallidum 14 3rd-Ventricle -> FS_nongm_vent 15 4th-Ventricle -> FS_nongm_vent 16 Brain-Stem 17 Left-Hippocampus 18 Left-Amygdala
24 CSF -> FS_nongm_csf
26 Left-Accumbens-area
28 Left-VentralDC
30 Left-vessel -------> *should be in list of FS_nongm? 31 Left-choroid-plexus -> FS_nongm_vent
41 Right-Cerebral-White-Matter -> FS_nongm_wm
43 Right-Lateral-Ventricle -> FS_nongm_vent 44 Right-Inf-Lat-Vent -> FS_nongm_vent
46 Right-Cerebellum-White-Matter -> FS_nongm_wm 47 Right-Cerebellum-Cortex
49 Right-Thalamus-Proper 50 Right-Caudate
52 Right-Pallidum 53 Right-Hippocampus 54 Right-Amygdala
58 Right-Accumbens-area
60 Right-VentralDC
62 Right-vessel -------> *should be in list of FS_nongm? 63 Right-choroid-plexus -> FS_nongm_vent
72 5th-Ventricle -> FS_nongm_vent
77 WM-hypointensities -> FS_nongm_wm
80 non-WM-hypointensities -------> *should be in list of FS_nongm_wm?
85 Optic-Chiasm -------> *should be in list of FS_nongm?
251 CC_Posterior -> FS_nongm_wm 252 CC_Mid_Posterior -> FS_nongm_wm 253 CC_Central -> FS_nongm_wm 254 CC_Mid_Anterior -> FS_nongm_wm 255 CC_Anterior -> FS_nongm_wm
It seem that the new version of mri_binarize that you had sent me might still include those four flagged regions in the 'mri_binarize --gm' output.
B) For the ROIs specific to the 2000aparc+aseg, I think it also seems pretty clear:
- 1000 and 2000 are returned as "ctx_*_unknown";
- All ROIs in the intervals [1001, 1003] and [1005, 1035] can be
returned, and these are all LH GM;
- All ROIs in the intervals [2001, 2003] and [2005, 2035] can be
returned, and these are all RH GM;
- ROIs '1004' and '2004' are the "ctx-?h-corpuscallosum", which would be
WM, but they don't seem to be returned.
C) For the ROIs specific to 2009aparc+aseg, I think it also seems pretty clear:
- 11100 and 12100 are returned as "ctx_*_Unknown";
- All ROIs in the intervals [11101, 11141] and [11143, 11175] can be
returned, and these are all LH GM;
- All ROIs in the intervals [12101, 12141] and [12143, 12175] can be
returned, and these are all RH GM;
- ROIs '11142' and '12142' are the "ctx_?h_Medial_wall", which would be
non-GM, but they don't seem to be returned.
D) And you mentioned that the regions in the interval [3000, 4035] are WM parcellations.
So, in total, does that seem like a complete categorization of regions that are typically returned, as well as their classification as either GM or non-GM? And in particular, I would be happy to know your thoughts on the flagged ROIs in the interval [0, 256].
Thanks, pt