Hello Shady,
The first thing to check is that the s/q-forms are set correctly for the images in question, if they are not then establihsing “true” left/right will be extremely difficult. Assuming these are set correctly, then all your images are consistent ( apart from the interim image between fslswapdim and fslorient ).
Confusion might possibly arise from that fact the the landmarks are on different sides of the head - in image1 the landmark is on the (anatomical) right, while for image2 it is on the left. Slices will always try to display images radiologically, you can force an image to be displayed “neurologically" ( e.g. by fslorient -x y z ) but this should be done with great caution as the orientation marker in the bottom corner of the slices output will be incorrect and most importantly the orientation information in the header will be wrong, any analysis/processing should be performed on the original, correct image.
Hope this helps
Kind Regards
Matthew
> On 28 Jun 2016, at 13:24, Shady Rahayel <[log in to unmask]> wrote:
>
> Hi,
>
> I have searched for previous threads about this topic and found several answers that have been beneficial to my understanding of image orientation. However, I'm still unsure about how I should handle the problem I have right now...
>
> I'm working with 2 datasets coming from 2 different labs. I've converted the dicom files from both labs to T1s and both are oriented in different ways, according to how the image was acquired in the scan. I've avoided using fslswapdim and fslorient and rather used fslreorient2std to rotate the image in the template orientation, as repeated many times before.
>
> For the first lab, all images had correct labels in FSLView. Using fslreorient2std, the images became oriented in the MNI152 orientation. I've used slicesdir `imglob *` to visualize in a single page the transformations that have happened. I've attached an example of the raw nii image and the reoriented nii image at this link: https://s32.postimg.org/st69ysn1h/Screenshot_Lab1.png
> And I notice that the patient's right hemisphere in the raw nii image now appears on the right side of the image in the reoriented nii image, which fits with the fslorient that says "neurological".
>
> For the other lab, all images also had correct labels and I have also used fslreorient2std. All images became reoriented correctly with the MNI152. However, when I used slicesdir `imglob *` to visualize images, I now notice that the patient's right hemisphere now appears on the left side of an image, which fits the fslhd saying it's in radiological convention. As an example, I've attached an example using anatomical landmarks that differ between both hemispheres at the link: https://s31.postimg.org/dzw64b51n/Screenshot_Lab2.png
> In the raw image, I used a sulcus as an anatomical landmark (ie, in the right hemisphere it's a straight line, whereas it's a hook-shaped sulcus). Even though I have read that radiological or neurological conventions do not matter for analysis inside FSL, I would like all images to be oriented neurologically like in the first lab (as we may use tools outside FSL). I have used the suggested commands: fslswapdim [reoriented image] -x y z [swapped image], which swapped the brain and gave what I wanted (ie, the line-shaped sulcus on the right side of the image and the hook-shaped sulcus on the left side of the image), although still having a fslorient saying that it is radiological. Then I used fslorient -swaporient [swapped image] to change the header and give a "radiological" convention to the image. However, as you can see from the image, it swapped back the image with the patient's right hemisphere on the left side of the image and vice versa. I understood that using both commands at the same time was about keeping the brain exactly as it was but changing the convention header only. In my case, I would like the patient's right hemisphere to be on the right side of the image, as with the first lab. What exactly should I be doing for this to happen? Only use fslswapdim [input] -x y z [output] and not caring about the header saying it's radiological?
>
> Thank you so much and sorry for such a long post,
>
> Shady
|