Michael Harms wrote:

Hi Doug,

Is --reorder4 a new feature in one of the more recent releases? (Currently, I'm using mri_convert from v4.1.0, and I don't see a -- reorder4 flag).

Yes, it is. It is actually only in the dev version.

Also, can you please clarify what do you mean when you said that -- out_orientation should not resample?

It should only change the geometry in the header, not the pixel values. What's your cmd?

If I take an nii file with this transform:

Orientation : PSR Primary Slice Direction: sagittal voxel to ras transform: -0.0229 -0.0300 1.1991 -88.7736 -0.9978 0.0634 -0.0251 144.1823 0.0628 0.9975 0.0377 -139.2478 0.0000 0.0000 0.0000 1.0000

and run it through "mri_convert --out_orientation LAS input.nii output.nii", then mri_convert spits out the following:

Setting output orientation to LAS Reslicing using trilinear interpolation

And the output.nii has this transform:

Orientation : LAS Primary Slice Direction: axial

voxel to ras transform: -1.0000 0.0000 0.0000 120.5695 0.0000 1.0000 0.0000 -97.4406 0.0000 -0.0000 1.2000 -97.0175 0.0000 0.0000 0.0000 1.0000

That is, both the "Reslicing using trilinear interpolation" comment and the xforms themselves appear to indicate that resampling was performed.

thanks for clarifying, -MH

On Fri, 2010-02-05 at 14:03 -0500, Douglas N Greve wrote:

...

Sorry, I don't have anything that will reorder and change the geometry to be correct. You might be able to do it with --reorder4 (and not --reorder) in combination with --out_orientation. --out_orientation should not resample.

doug

Michael Harms wrote:

...

Hello Doug, Is it possible using mri_convert to just reorder the way in which the data is stored, without any resampling of the actual voxel values?

i.e., I have Siemen's MPRAGE DICOMs, and if run mri_convert, I get a default "PIL" orientation. I would like to just reorder the way in which the data is stored, to get a "LAS" orientation instead.

I tried using "--out_orientation LAS", but that does a resampling (interpolation) such that the resulting xform has strictly 0's on the off-diagonal. And reading the mri_convert help, it clearly states that the --out_orientation flag is to be used when the input orientation is WRONG, which is not the case here (since I presume that the input orientation is correctly read out of the DICOMs).

I tried "--reorder 2 1 3", to see what effect that has on a file already in nii format, but I got the message: in nii format, but I got the message: reordering axes... MRIreorder: incorrect dst width error reordering axes

I've checked the archives, and read through the mri_convert help, but just can't seem to figure out if reordering from one orientation to another (without resampling) is possible using mri_convert.

thanks, -MH

It may be that tkmedit is doing the interpolation. Try running mri_diff with the --pixonly option. This just tests for differences in pixel value, excluding the geometry, etc.

Michael Harms wrote:

Command (FS v4.1.0) was just: mri_convert --out_orientation input.nii output.nii

Then, loading both input.nii and output.nii into tkmedit, it is clear that the voxel values at the same cursor location do not match. (output.nii is clearly smoother, consistent with it being interpolated).

thanks, -MH

On Fri, 2010-02-05 at 14:30 -0500, Douglas N Greve wrote:

...

Michael Harms wrote:

...

Hi Doug,

Is --reorder4 a new feature in one of the more recent releases? (Currently, I'm using mri_convert from v4.1.0, and I don't see a -- reorder4 flag).

Yes, it is. It is actually only in the dev version.

...

Also, can you please clarify what do you mean when you said that -- out_orientation should not resample?

It should only change the geometry in the header, not the pixel values. What's your cmd?

...

If I take an nii file with this transform:

Orientation : PSR Primary Slice Direction: sagittal voxel to ras transform: -0.0229 -0.0300 1.1991 -88.7736 -0.9978 0.0634 -0.0251 144.1823 0.0628 0.9975 0.0377 -139.2478 0.0000 0.0000 0.0000 1.0000

and run it through "mri_convert --out_orientation LAS input.nii output.nii", then mri_convert spits out the following:

Setting output orientation to LAS Reslicing using trilinear interpolation

And the output.nii has this transform:

Orientation : LAS Primary Slice Direction: axial

voxel to ras transform: -1.0000 0.0000 0.0000 120.5695 0.0000 1.0000 0.0000 -97.4406 0.0000 -0.0000 1.2000 -97.0175 0.0000 0.0000 0.0000 1.0000

That is, both the "Reslicing using trilinear interpolation" comment and the xforms themselves appear to indicate that resampling was performed.

thanks for clarifying, -MH

On Fri, 2010-02-05 at 14:03 -0500, Douglas N Greve wrote:

...

Sorry, I don't have anything that will reorder and change the geometry to be correct. You might be able to do it with --reorder4 (and not --reorder) in combination with --out_orientation. --out_orientation should not resample.

doug

Michael Harms wrote:

...

Hello Doug, Is it possible using mri_convert to just reorder the way in which the data is stored, without any resampling of the actual voxel values?

i.e., I have Siemen's MPRAGE DICOMs, and if run mri_convert, I get a default "PIL" orientation. I would like to just reorder the way in which the data is stored, to get a "LAS" orientation instead.

I tried using "--out_orientation LAS", but that does a resampling (interpolation) such that the resulting xform has strictly 0's on the off-diagonal. And reading the mri_convert help, it clearly states that the --out_orientation flag is to be used when the input orientation is WRONG, which is not the case here (since I presume that the input orientation is correctly read out of the DICOMs).

I tried "--reorder 2 1 3", to see what effect that has on a file already in nii format, but I got the message: in nii format, but I got the message: reordering axes... MRIreorder: incorrect dst width error reordering axes

I've checked the archives, and read through the mri_convert help, but just can't seem to figure out if reordering from one orientation to another (without resampling) is possible using mri_convert.

thanks, -MH