As an update, I noticed that the bbregister output log had an interesting warning statement:
WARNING:: Flipping Left/Right orientation (as det < 0) tkregister_tcl /Applications/freesurfer/tktools/tkregister2.tcl target volume /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/tmp.bbregister.4857/fslregister/movvol.fslregister.nii movable volume /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/tmp.bbregister.4857/template.nii reg file /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/tmp.bbregister.4857/fslregister/swapreg.dat LoadVol 0 ZeroCRAS 0 $Id: tkregister2.c,v 1.121.2.1 2011/03/28 20:25:16 greve Exp $ Diagnostic Level -1 INFO: loading target /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/tmp.bbregister.4857/fslregister/movvol.fslregister.nii INFO: target does not conform to COR format, so I'm going to reslice to COR. This will not affect the final registration.
Not sure if this is common. I'll e-mail back if I figure anything else out.
MP
On Tue, Aug 6, 2013 at 9:13 AM, Mark Plantz < markplantz2016@u.northwestern.edu> wrote:
I think that the registration turned out ok [well it isn't upside down]. I attached a few screenshots.
My bbregister command is:
*bbregister --s CMH-8169 --mov /Volumes/Wong_Lab/CochlearImplant/UNC_NiFTI/infant-1yr-mgz/seg.mgz --init-fsl --reg /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/register.dat --t1 --o /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/output.mgz*
Subsequently, my mri_label2vol command is:
*ingvalsons-imac:~ IngvalsonLab$ mri_label2vol --seg /Volumes/Wong_Lab/CochlearImplant/UNC_NiFTI/infant-1yr-mgz/seg.mgz --temp /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/output.mgz --reg /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/register.dat --o /Users/IngvalsonLab/Desktop/output.mgz *
Number of labels: 0 Annot File: (null) Template Volume: /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/output.mgz Outut Volume: /Users/IngvalsonLab/Desktop/output.mgz Registration File: /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/register.dat Fill Threshold: 0 Label Vox Vol: 1 ProjType: (null) ProjTypeId: 0 ProjStart: 0 ProjStop: 0 ProjDelta: 0.1 Subject: (null) Hemi: (null) UseNewASeg2Vol: 1 DoLabelStatVol 0 LabelCodeOffset 0 setenv SUBJECTS_DIR /Applications/freesurfer/subjects $Id: mri_label2vol.c,v 1.34.2.5 2012/06/08 17:31:03 greve Exp $ Template RAS-to-Vox: -------- -1.000 0.000 0.000 128.000; -0.000 -0.000 -1.000 128.000; -0.000 1.000 -0.000 128.000; 0.000 0.000 0.000 1.000; Template Voxel Volume: 1 nHits Thresh: 0 Loading registration from /Volumes/Wong_Lab/LabMRIData/FreeSurfer/CMH-8169/bbregister/register.dat RegMat: -------- -0.994 -0.098 0.042 0.715; 0.043 -0.001 0.999 -14.682; 0.097 -0.995 -0.005 2.026; 0.000 0.000 0.000 1.000; Label RAS-to-Vox: -------- 0.994 0.098 -0.042 127.285; -0.097 0.995 0.005 125.974; 0.043 -0.001 0.999 113.318; 0.000 0.000 0.000 1.000; ASeg2Vol: Building LUT ASeg2Vol: Sorting ASeg2Vol: Mapping ASeg2Vol: Reverse Map nmisses = 9739499 (73560 filled) ASeg2Vol: done ingvalsons-imac:~ IngvalsonLab$
Is it significant that number of labels = 0? Or is that because I am using a segmentation volume instead of an actual .label file?
Other than that, I cannot seem to determine why the output images would be inverted by 180 degrees.
Thank you in advance MP
On Mon, Aug 5, 2013 at 4:26 PM, Douglas Greve greve@nmr.mgh.harvard.eduwrote:
did you look at the registration with tkregister2? When bbregister finished, it would have printed out the command line to run (it is also at the end of the log file). If that looks ok, please send your bbregister command. doug
On 8/5/13 1:47 PM, Mark Plantz wrote:
Thanks Doug. mri_label2vol worked well, except it looks like all of the files are inverted by 90 degrees. They appear inverted in both tkmedit and 'FreeView.' I read that inversions are common in "fslview" but not in tkmedit. Any idea what could be causing this problem?
My command line is:
*mri_label2vol --seg /Volumes/Wong_Lab/CochlearImplant/UNC_NiFTI/infant-1yr-mgz/seg.mgz --temp /Users/IngvalsonLab/Desktop/temp.mgz --o seg_new.mgz --reg register.dat*
I didn't encounter any errors with mri_label2vol nor bbregister [which was used to generate register.dat and temp.mgz].
Any ideas what could be causing this problem?
I can attach more images if this one doesn't portray the inversion well.
Thanks a lot
MP
On Mon, Aug 5, 2013 at 10:12 AM, Douglas Greve <greve@nmr.mgh.harvard.edu
wrote:
The file that you pass it with the --reg option is the registration file (eg, --reg register.dat). This can be used for input to mri_vol2vol or mri_label2vol. mri_label2vol is appropriate for a segmetation
doug
On 8/5/13 10:22 AM, Mark Plantz wrote:
Thank you. Do you know how to access the transform that is outputted by bbregister. I am just obtaining an output volume <volume.mgz>>?
On Thu, Aug 1, 2013 at 3:56 PM, Bruce Fischl < fischl@nmr.mgh.harvard.edu> wrote:
Hi Mark
you can use the transform that bbregister outputs and give it to mri_convert with -at <transform name>
cheers
Bruce
On Thu, 1 Aug 2013, Mark Plantz wrote:
One last question: Is there any flag I can use with 'bbregister' that
prevents the labels from changing in my segmentation volume? This is occuring because the volume that I am moving has labels. For some reason, the labels change ever so slightly in the output file.
On Thu, Aug 1, 2013 at 2:28 PM, Bruce Fischl < fischl@nmr.mgh.harvard.edu> wrote: glad it worked out Bruce On Thu, 1 Aug 2013, Mark Plantz wrote:
Just as an update, it turns out that my segmentation volume had been slightly altered during the conversion from .nii to .mgz. I completely forgot to use the -rt nearest flag to prevent the labels from becoming distorted. That is why the segmentation looked so fuzzy with various regions labeled incorrectly. Apparently, I should check the basics before overcomplicating the problem, haha. Thanks for the help everyone! On Thu, Aug 1, 2013 at 9:36 AM, Mark Plantz <markplantz2016@u.northwestern.edu> wrote: Sounds good. I also just realized that the segmentation values are varying from 0-255, which makes me think that the file is only designed for the grayscale/heat/etc. settings [where 0 is darkest regions and 255 is brightest regions]. Is that a valid assumption? Also, if this is the case, would it be impossible to create a .gca file using mri_ca_train with these files? Thanks for the help. MP On Wed, Jul 31, 2013 at 8:43 PM, Douglas Greve <greve@nmr.mgh.harvard.edu> wrote: Just make sure that you create a new file and do not repeat index numbers. On 7/31/13 2:02 PM, Mark Plantz wrote: That makes sense. Wouldn't that cause there to be some overlap between different regions, since the same index #'s used by the two programs (FreeSurfer & MRIcro) correspond to different brain regions? On Tue, Jul 30, 2013 at 12:50 PM, Douglas Greve <greve@nmr.mgh.harvard.edu> wrote: for every line in that file, you will need to create a line in the new LUT, something like 8 Frontal_Mid_R 0 255 0 0 This will make the right mid frontal green = (0,255,0) On 7/30/13 9:55 AM, Mark Plantz wrote: Hmm, so I received the text part of the LUT. Looks like this is completely different from the segmented regions that we would want. [attached]. Can't seem to understand how this text file could be related to the default FreeSurfer LUT. It looks like the labeling for an AAL map. This probably wont' work will it? - Mark On Tue, Jul 30, 2013 at 8:47 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: the UNC one? No, you should ask them On Tue, 30 Jul 2013, Mark Plantz wrote: ---------- Forwarded message ---------- From: Mark Plantz <markplantz2016@u.northwestern.edu> Date: Tue, Jul 30, 2013 at 8:39 AM Subject: Re: [Freesurfer] infant atlas segmentation To: Bruce Fischl <fischl@nmr.mgh.harvard.edu> Just opened up the default FreeSurferColorLUT.txt file. So it looks like all I have to change is the #No. column of the file, to match the ones in the other LUT used. I actually just received the MRIcro .lut file from the UNC group. [attached] Hopefully this will help with the conversion. However, I still have to figure out how to open the file. Any ideas? On Tue, Jul 30, 2013 at 8:25 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: Hi Mark creating a FreeSurfer LUT really isn't that hard. Have you looked at our standard one? Probably we have the names you already want and you just need to change our indices to match the ones in your segmentation file (and save it as a different LUT) cheers Bruce On Tue, 30 Jul 2013, Mark Plantz wrote: Turns out that the files had been labeled using a color LUT in a program called MRIcro. I am now installing this program simply to access the LUT. Will it be as simple as exchanging the index values from this LUT to a custom LUT in FreeSurfer? On Tue, Jul 30, 2013 at 12:35 AM, Douglas Greve <greve@nmr.mgh.harvard.edu> wrote: The problem is that you need to create a completely new LUT. The labels that get displayed with the standard LUT are going to be totally random with respect to what they should display. On 7/29/13 2:20 PM, Mark Plantz wrote: Hi Doug, Thanks for the reply. I will definitely go into the LUT and change those values to some specific region. My main concern is that I think all of these "out of bounds" labels are the symptoms of some larger problem. The segmentation looks very strange as compared to a typical segmentation (i.e. an 'aseg.mgz' file). In fact, the segmentation appears to be labeling "left-<region" in the right hemisphere. Have you seen a segmentation file that looks like the once I attached before? I am used to seeing segmentation files such as the one attached to this e-mail. - Mark On Mon, Jul 29, 2013 at 12:28 PM, Douglas Greve <greve@nmr.mgh.harvard.edu> wrote: Mark, the only thing that is wrong with the out of bounds regions is that they are not represented in the LUT. Once you add an index corresponding to the index of the segmentation to the LUT, it will not show up as "out of bounds". The brain mask has nothing to do with it. best doug On 7/29/13 12:50 PM, Mark Plantz wrote: Sorry about that. I received the manual segmentations from aUNCmedicalgroup<http://www.med.unc.edu/bric/ideagroup/free-softwares/unc-in fan t-0-1-2 -atla ses>. I actually asked the main researcher about the out of bounds regions. He thinks that as long as the segmented file lines up with the input volumes, a few out of bounds regions might be OK. However, these seg volumes look very different from a typical .aseg volume that I am used to seeing after recon-all is run. Maybe the segmentation is simply incompatible with FreeSurfer?
Thanks for the help. These files are a nightmare! - MP On Mon, Jul 29, 2013 at 11:40 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: can you cc the list please? Maybe we should start over. Where did you get the manual segmentations from? On Mon, 29 Jul 2013, Mark Plantz wrote: Ahh, right. So it looks like the red regions appear to be averaged around 248. I am not sure why that is the case or if it has any significance. I attached a screen shot for reference. On Mon, Jul 29, 2013 at 11:29 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: it isn't a registration file. You should just specify it as a volume on the freeview command line, the same as brainmask.mgz On Mon, 29 Jul 2013, Mark Plantz wrote: I actually tried to view the brainmask.mgz file with the segmentation file [Freeview calls it 'registration file' I believe]. For some reason, the file will load without segmentation, but when I attempt to add the registration file, I get an error message: "Failed to load MRI ~/.../../../brainmask.mgz Could this potentially be part of the problem? Thanks MP On Mon, Jul 29, 2013 at 10:57 AM, Mark Plantz <markplantz2016@u.northwestern.edu> wrote: It looks like the values vary from 20-120 depending on which region I run the cursor over. However, aren't those the values that correspond to the brain.mgz regions? Is there another way to find the segmentation values? On Mon, Jul 29, 2013 at 10:37 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: I mean the segmentation values that give you the "out of bounds" message On Mon, 29 Jul 2013, Mark Plantz wrote: Hi Bruce, Do you mean the brainmask.mgz values? I'm not sure what the segmentation #'s exactly are. Thanks! MP On Mon, Jul 29, 2013 at 10:17 AM, Bruce Fischl <fischl@nmr.mgh.harvard.edu> wrote: Hi Mark what #s are those segmentations? It should be trivial to just add them to the end of the LUT. I wouldn't let them be out of bounds as there might be code around that might ignore them then. cheers Bruce On Mon, 29 Jul 2013, Mark Plantz wrote: Hi Doug, Thanks for the reply. After doing some research, it looks like messing with the LUT might be a risky decision [and way out of my abilities!]. If I were to use this segmentation file to create a .gca atlas using the command mri_ca_train, do you think it would yield problems or b...
[Message clipped]