Dear Martin and all,
I had actually checked the RAS coordinates for all of my subjects and timepoints, and the one I mentioned is the only one in which the rotation is present.
I created a uchar image from rawavg.mgz and tried to run recon-all from that without the motioncor step and without talairach (creating talairach.xfm manually from scratch using tkregister2), but recon-all crashes during mri_nu_correct (see error below).
One strange observation: I had been running this subject in version 5.0 previously and noticed that, although the RAS coordinates and transform are the same as the ones I get in version 5.1, I do not see any smoothing effect when comparing rawavg.mgz and orig.mgz.
I guess the best we can do at this stage is to manually edit the input nii file so that no rotation is coded in there (by changing the sform/qform as suggested by Michael Harm) - unless you have other suggestions...
Thanks very much again for your help, Irene
$Id: mri_convert.c,v 1.179.2.2 2011/05/16 20:53:47 greve Exp $ reading from ./tmp.mri_nu_correct.mni.22469/nu2.mnc... TR=0.00, TE=0.00, TI=0.00, flip angle=0.00 i_ras = (0, -1, 0) j_ras = (-0.0436194, 0, 0.999048) k_ras = (0.999048, 0, 0.0436194) INFO: transform src into the like-volume: orig.mgz writing to nu.mgz... mri_make_uchar nu.mgz transforms/talairach.xfm nu.mgz MRIvoxelToTalairachVoxel: unsupported slice direction 1 MRIvoxelToTalairachVoxel: unsupported slice direction 1 Numerical result out of range MRIvoxelToTalairachVoxel: unsupported slice direction 1 Numerical result out of range MRIvoxelToTalairachVoxel: unsupported slice direction 1 Numerical result out of range
Le vendredi 04 novembre 2011 à 16:17 -0400, Martin Reuter a écrit :
Irene,
I don't really know why the RAS coordinates are slightly rotated. I checked and they are orthogonal. I don't know where this comes from (autoalign on the scanner, but then why would it be only in one of your volumes).
Maybe Doug knows a trick how to avoid the interpolation as it seems unnecessary?
Also, when you say you checked the other subject, did you look at the images, or did you check the RAS coordinates in the mri_info output. It can be that rotation is small and you won't see it the image but it is still there. The RAS xform x_r,x_a,x_s, y_r, y_a, y_s and z_r z_a z_s should only be +-1 or zero.
Best, Martin
On Fri, 2011-11-04 at 17:46 +0100, Irene.Altarelli@ens.fr wrote:
Dear Martin,
I have a single scan per session, so no averaging. Also, this is the only image in my longitudinal data for which I am encountering this problem. I have checked all other subjects and none of them is rotated/smoothed.
I am not sure I understand precisely why this scan needs to be rotated, is it to establish RAS coordinates' center and axes? Do you think there is a way I can fix it? I don't have many subjects in this protocol, so dropping one would be a real pity.
Thanks very much again for your help. Best, Irene
Selon Martin Reuter mreuter@nmr.mgh.harvard.edu:
Hi Irene,
the voxel sizes are isotropic 1mm (which is good). However for some reason the first image is slightly rotated while the second is perfectly aligned with the axes. Therefore the first gets interpolated while the second stays the same. That explains the different smoothing.
Now the question is why is one image different. Do you collect a single scan for each session or are you averaging several within-session scans? You should look at other subjects and try to find out if and why specific time points are slightly rotated.
By the way this is a good example how to potentially mess up a longitudinal study. If for example one of the time points is always (or more frequently) rotated because of a different protocol the study would be severely biased. So it is good that you are checking this.
Best, Martin
On Fri, 2011-11-04 at 10:28 +0100, Irene.Altarelli@ens.fr wrote:
Dear Martin,
the RAS coordinates and voxel to ras transforms are indeed slightly
different.
Should I change this and if so, how?
Please find the output of mri_info for each of the two acquisitions at the
end
of this email.
Thanks again! Irene
Volume information for gsTP1/mri/rawavg.mgz type: MGH dimensions: 256 x 256 x 176 voxel sizes: 1.0000, 1.0000, 1.0000 type: FLOAT (3) fov: 256.000 dof: 0 xstart: -128.0, xend: 128.0 ystart: -128.0, yend: 128.0 zstart: -88.0, zend: 88.0 TR: 0.00 msec, TE: 0.00 msec, TI: 0.00 msec, flip angle: 0.00 degrees nframes: 1 PhEncDir: UNKNOWN ras xform present xform info: x_r = 0.0000, y_r = -0.0436, z_r = 0.9990, c_r =
-3.5493
: x_a = -1.0000, y_a = 0.0000, z_a = 0.0000, c_a =12.8739
: x_s = 0.0000, y_s = 0.9990, z_s = 0.0436, c_s =-16.0490
talairach xfm : Orientation : PSR Primary Slice Direction: sagittal
voxel to ras transform: 0.0000 -0.0436 0.9990 -85.8823 -1.0000 0.0000 0.0000 140.8739 0.0000 0.9990 0.0436 -147.7657 0.0000 0.0000 0.0000 1.0000
voxel-to-ras determinant -1
ras to voxel transform: -0.0000 -1.0000 -0.0000 140.8739 -0.0436 0.0000 0.9990 143.8789 0.9990 -0.0000 0.0436 92.2460 0.0000 0.0000 0.0000 1.0000
Volume information for gsTP2/mri/rawavg.mgz type: MGH dimensions: 256 x 256 x 176 voxel sizes: 1.0000, 1.0000, 1.0000 type: FLOAT (3) fov: 256.000 dof: 0 xstart: -128.0, xend: 128.0 ystart: -128.0, yend: 128.0 zstart: -88.0, zend: 88.0 TR: 0.00 msec, TE: 0.00 msec, TI: 0.00 msec, flip angle: 0.00 degrees nframes: 1 PhEncDir: UNKNOWN ras xform present xform info: x_r = 0.0000, y_r = 0.0000, z_r = 1.0000, c_r =
-1.5339
: x_a = -1.0000, y_a = 0.0000, z_a = 0.0000, c_a =11.5254
: x_s = 0.0000, y_s = 1.0000, z_s = 0.0000, c_s =-19.3390
talairach xfm : Orientation : PSR Primary Slice Direction: sagittal
voxel to ras transform: 0.0000 0.0000 1.0000 -89.5339 -1.0000 0.0000 0.0000 139.5254 0.0000 1.0000 0.0000 -147.3390 0.0000 0.0000 0.0000 1.0000
voxel-to-ras determinant -1
ras to voxel transform: -0.0000 -1.0000 -0.0000 139.5254 -0.0000 -0.0000 1.0000 147.3390 1.0000 0.0000 0.0000 89.5339 0.0000 0.0000 0.0000 1.0000
Selon Martin Reuter mreuter@nmr.mgh.harvard.edu:
Hi Irene,
since those images are not from the longitudinal runs, it really is a cross sectional question:
You have two images and the raw average look both similarly noisy, but the orig (which is the first image in the recon-all stream) is smoother for one than the other.
That means that the conform step (that interpolates the isotropic orig from the rawavg) introduces the different smoothing. Not sure why it is different. You can check (with mri_info) the voxel sizes of your rawavg and the RAS coordinates. If they are the same, smoothing should also be the same.
Best, Martin
On Thu, 2011-11-03 at 17:48 +0100, Irene Altarelli wrote:
Dear Martin,
thanks for your quick reply.
I am using version 5.1. I only have two timepoints per subject and I took the snapshots from the independent runs. What bothers me is that the difference is not present in the 001.mgz images, but seems to
appear
later in the workflow.
Thanks again, Irene
Le jeudi 03 novembre 2011 à 10:40 -0400, Martin Reuter a écrit : > Hi Irene, > > both images should be smoothed the same. You only have two time
points
> in each subject? Which FreeSurfer version are you using? And these > images are from the *.long.base directories (or are they from the > independent runs)? > > -Martin > > On Thu, 2011-11-03 at 15:20 +0100, Irene Altarelli wrote: > > Dear Fs experts, > > > > I am running a longitudinal dataset through the longitudinal
pipeline,
> > and noticed that for one subject, although we kept everything the
same
> > (sequence, coil etc) the two timepoints' brainmasks look different: > > timepoint one (in the attached snapshot, gs_tp1) looks as if it has
been
> > smoothed, compared to timepoint two (gs_tp2). I have checked some
other
> > images from this subject, and the same seems to occur for orig.mgz,
but
> > not for rawavg.mgz. > > > > Any idea on what could be causing the difference would be greatly > > appreciated. > > > > Thanks in advance, > > Irene > > _______________________________________________ > > Freesurfer mailing list > > Freesurfer@nmr.mgh.harvard.edu > > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > _______________________________________________ > 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
> 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
-
Irene Altarelli