Hi,

I have thought this for a long time, but I still can not understand this. I have some questions about fslswapdim

(1) Take an easy example: If only one voxel in the mask and the coordinate is -1 1 1, when I used fslswapdim orig_mask -x y z output_mask, the coordinate of the output_mask would become 1 1 1, right?

No, not really.  What fslswapdim does is to swap the data around.  However, if you are looking at one particular point of the anatomy then that is equivalent to its coordinate changing.  It is the voxel coordinates that are swapped though, so there is no guarantee that -1 1 1 would become 1 1 1 with a -x y z swap unless the mm coordinates (as these must be mm coordinates, since voxel coordinates are always positive) are perfectly arranged to have x=0 at the exact centre of the volume.

(2) You agreed with Sourena's method, right? 

fslswapdim orig_t1 -x y z flipped_t1

fslswapdim orig_mask -x y z flipped_mask

Yes, in the sense that this will flip the general sense of left and right, but it won't generally create images which can be neatly overlaid at this point.  That requires an extra registration step.

Use flirt with 6DOF to register flipped_t1 into orig_t1, and apply the transformation matrix to your flipped_mask.

A. Specifically, what's the orig_t1? The mask is from the group level results which have been registered to MNI152_T1_2mm_brain.nii.gz, so the orig_t1 is MNI152_T1_2mm_brain.nii.gz?

I believe that we were talking about the orig_t1 being a T1 structural image, not in standard space (as you wrote about "the T1 image" which sounded more like a structural image of a single individual rather than an image in  standard space image).  

If all your images are in standard space then this is a special case and not what I think we were talking about to start with.  Only with the standard space images can you avoid this registration step because in this case the anatomy is aligned with the axes and x=0 is exactly at the centre of the volume.  So only in this case does the fslswapdim command with -x y z produce a flipped image that requires no further processing.  

B. How to apply the transformation matrix to your flipped_mask? Should I use flirt or applywarp? What's the input and reference? Again, the mask has been registered to standard T1, why register it again？

Again, it was unclear before that everything was in standard space.  If so then you can just use fslswapdim and avoid the extra registration.

C. When you said "There's no reason that the mid-saggital plane of the T1 image should be aligned with the yz-plane at the mid-point on the x-axis.", you mean because the brain is not symmetrical? 

No, I was referring to the case of an individual structural image, prior to it being transformed into standard space.  When not in standard space then orientation of the head in the image is unlikely to be exactly aligned with the axes (even if the radiographer tries to set this up in the slice prescriptions).  Hence the flipped usually creates a change in the tilt or position of the head.  In standard space it does not and you do not need to worry about this.

All the best,

Mark

Thanks so much.

Mingxia

On Fri, May 10, 2013 at 9:54 AM, Mark Jenkinson <[log in to unmask]> wrote:

Hi,

There's no reason that the mid-saggital plane of the T1 image should be aligned with the yz-plane at the mid-point on the x-axis.  The chances are that this flip will also change the orientation and translation.  So you do need to do the registration.

All the best,

Mark

On 10 May 2013, at 07:19, zhang mingxia <[log in to unmask]> wrote:

Hi Sourena,

Thanks a lot. I tried:

 fslswapdim  orig_input -x y z flipped_output

Because the orig_input has been registered into the T1 image, I don't need to do further registration, right?

Mingxia

On Wed, May 8, 2013 at 8:48 AM, Sourena Soheili <[log in to unmask]> wrote:

Hi,
I have not done it before and others may have better ideas;but maybe it works:
swap left/right axis signs of both your original T1 and mask. If the
left/right axis is x it would be:
fslswapdim orig_t1 -x y z flipped_t1
fslswapdim orig_mask -x y z flipped_mask

Use flirt with 6DOF to register flipped_t1 into orig_t1, and apply the
transformation matrix to your flipped_mask.

Cheers,
Sourena

On 5/8/13, zhang mingxia <[log in to unmask]> wrote:
> Dear FSL experts,
>
> I have a result in left hemisphere and want to investigate the laterality
> of this area. So, I planned to make a mask of this area then make the
> corresponding right side of mask based on the first mask and then extract
> the mean value of the two masks for further calculation.
>
> My problem is I don't know how to make the corresponding right side of mask
> based on mask in the left hemisphere. Hope to get some advice.
>
> Thanks in advance.
>
> Mingxia Zhang
>