Hi,

Just to add to Paul’s email - it is also possible that what you are seeing is a registration failure, where it has incorrectly rotated the image by a large degree. This can result in labels being attached to the wrong parts of the anatomy, but is not really a failing in the labelling, or in any flipping, but just in the fact that the registration was poor.  Usually this can be fixed by using fslreorient2std followed by -nosearch on the flirt call.

As Paul says, flirt will *never* flip the left-right order (in terms of handedness) of the coordinate system, but that doesn’t mean that registrations never fail, and a full 180 rotation in various axes (which involves no flipping) will still end up giving you a result that looks very wrong, with labels that are also very wrong.  So I suspect that what you need is to fix the registration and not be concerned about the labels.

If you would prefer not to use fslreorient2std then you can also use -nosearch and -usesqform in the flirt call, as that should initialise the matrix to get the initial alignment correct.  Also, note that fslreorient2std does not flip left-right ordering either, but just does combinations of 90, 180 and 270 degree rotations about different axes.

All the best,
Mark 

On 25 Aug 2022, at 12:09, paul mccarthy <[log in to unmask]> wrote:

Hi Ruifeng,

FLIRT will preserve orientation labels when registering/transforming an image - if the orientation of the FLIRT/FAST output is incorrect, then this suggests that the orientation of the input is incorrect.

Are the orientation labels correct for both of your raw images? If not, then you need to fix this before performing any processing. You can find some information on this problem here: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Orientation%20Explained

Paul


On Tue, 23 Aug 2022 at 15:57, Ruifeng Dong <[log in to unmask]> wrote:
Hi all,

I am a new user of FIRST. 

I am trying to do brain segmentation using FIRST. My data came from two scanners. One uses LPI orientation and the other uses RAI. The subcortical structure volumes look fine for LPI images, but apparently wrong for RAI ones. See attached. Then I flipped the RAI images using fslreorient2std and the volumes look fine again. This seems ok, but how do we know whether the left and right are flipped during the registration, as they are roughly symmetric? (Question 1)
<image.png>

I then tried to dig deeper. I registered the two images using FLIRT and MNI152_T1_1mm.nii.gz. Again the AP and LR directions are wrong for the RAI images. See attached. Is this caused by the initial value of rotation matrix chosen in FLIRT? If so, can we just manually set it to solve this issue? (Question 2)
A slice of registered RAI image:
<image.png>

The slice in MNI152:
<image.png>

Thank you!

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1


To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1



To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1