Thanks for your explanation, John.

Firstly, I agree with you that the coreg in SPM (6dof) is a bit rough. It is one of the steps to find a TMS stimulating point from MNI space to individual space.

That is actually one of the reason we use flirt, which did 12 dof affine transform. So, svd(Xfm(1:3,1:3)) showed a scaling factor.

I think there is a good point in the part your cited in fsl website. It seems the centre of mass of the volume was used as the centre. So I might need to find the additional transform matrix to make the shift, which could be the last puzzle.

At the meantime, could you suggest how to do a 12dof affine "coreg" in spm? I'm thinking to re-do this step on SPM, if possible.

Thanks,

Chao

On 26 October 2017 at 21:03, John Ashburner <[log in to unmask]> wrote:

My first point is that you won't get very good alignment if you just use Coreg to get to MNI space, as it only does rigid alignment and doesn't handle the different sized and differently shaped brains.

I'm not quite sure what Mark J and the guys in FMRIB use for encoding their transforms, so I can't say much here.  What I do notice though is that the Xfm does not encode a rigid body transform. Ignoring the translations, the transform is decomposed into a rigid body rotation, some isotropic zooms and another rigid body transform, which can be computed via SVD.  The zooms are:

> svd(Xfm(1:3,1:3))

ans =

    1.0730

    1.0291

    0.9739

This suggests that the matrix does not encode a rigid transform.  If you called FLIRT with 6 dof, then I really have no idea what the matrix encodes.

I seems that FLIRT ignores things like the origin in the images, which may account for the very different translations.  According to https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FLIRT/FAQ :

The FSL convention for transformation matrices uses an implicit centre of transformation - that is, a point that will be unmoved by that transformation, which is an arbitrary choice in general. This arbitrary centre of the transformation for FSL is at the mm origin (0,0,0) which is at the centre of the corner voxel of the image.

When using the transformation parameters from FLIRT, there is an additional complication in that the parameters are calculated in a way that uses a different centre convention: the centre of mass of the volume. The effect of this is that each of the three matrices above end up with an adjustment in the fourth column (top three elements only) that represents a shift between the corner origin and the centre of mass, while the rest of the matrix (first three columns) is unaffected. Once that is done the matrices are multiplied together, as indicated above, and you get your final matrix.

Best regards,

-John

On 26 October 2017 at 01:37, Chao Suo <0000171bf1c6344d-dmarc-reques[log in to unmask]> wrote:

Dear SPM/FSL users,

I have a question about the inconsistent transform matrix during linear corregister using SPM and FSL.
For example, I have one image (Qform is A) coregistered to an MNI space using a standard MNI template using "Coregister: Estimate" function.
The new image’s Qform is A’. The transform matrix (TE_spm) my understand is equal to A’*inv(A).
I can use this matrix (TE_spm) to reorient the Image to the MNI space using the function Batch-> SPM-> Util -> reorient images

When I use “flirt” in FSL toolbox, the output image is also aligned to MNI space. However, the output matrix (xfm) is way different with the TE_spm I generate.
And it doesn’t work with the “reorient images” function in SPM. Further, as the function “flirt” will reslice (resample) the image by default, the Qform of A’ generated by flirt is re-set to the centre of the image matrix.

My question is how to apply the xfm to reorient the input image A to MNI space? Also, is it possible to convert the xfm in FSL to the matrix that can be used in reorient image function?
I have attached the matrix below, for more information.


A=[ -0.9992   -0.0244    0.0314   97.7589
   -0.0244    0.9997    0.0008 -133.7938
    0.0314   -0.0000    0.9995 -149.3965
         0         0         0    1.0000]

A’=[   -0.9995   -0.0092    0.0317   93.3496
    0.0067    0.8845    0.4665 -176.7222
    0.0323   -0.4665    0.8839  -58.0182
         0         0         0    1.0000]

TE_spm= [    0.9999    0.0152    0.0003   -2.3183
   -0.0136    0.8844    0.4665   12.6318
    0.0069   -0.4665    0.8845   11.0441
         0         0         0    1.0000 ]

Xfm = [1.015110314  0.001403876797  -0.03273563415  -2.95397524
-0.0340642576  0.881267329  0.4458567806  -42.9856035
0.03133599993  -0.4915636726  0.9529941246  7.78935539
0  0  0  1 ]


Thanks,

Chao

--

Prof John Ashburner
Professor of Imaging Science
UCL Institute of Neurology
Queen Square
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square, London, WC1N 3BG
E: [log in to unmask]  T: +44 (0)20 3448 4365
http://www.fil.ion.ucl.ac.uk/