2010/6/15 Mark Jenkinson <[log in to unmask]>

Dear Clark,

I am sorry that the FAQ was out of date on this.
I have just modified the FAQ to adjust it for FSL4.1.
Basically, it no longer matters whether the storage is
"neurological" or "radiological" as the FSL tools will
work correctly with either, or even with a mix of them.
This did not use to be the case but it is now.

As for your inconsistent qform and sform - THIS IS
EXTREMELY BAD! You should convert your images
again from the original DICOM, as if you have such
inconsistent information then there is NO WAY of
knowing which is the correct left or right in the image,
since the qform information says one thing and the
sform information says another. This is why we have
the warnings. We also use dcm2nii for our conversions
but have not noticed this problem. However, we do not
use the "reoriented" images. So check whether the other
images have consistent information in the qform and sform
(the sign of the determinant of these matrices needs to be
the same). Just running fslorient will tell you what the
situation is.

With respect to FSLView, it is the case that it tries to display
the images in a "radiological display" although this doesn't
have an obvious meaning for images that are differently
oriented to the standard MNI152. It is not a problem for
any of the tools how things are shown in FSLView, as all
tools will work correctly for any storage type now. If you
do wish to see the images in the same orientation as the
MNI152 template images then you should run the command:

fslswapdim invol RL PA IS outvol
and if it gives you the message

Cannot perform requested swap (NERUOLOGICAL/RADIOLOGICAL storage inverted)
Try the following command instead
Fslswapdim o_dcm2nii_Output.nii.gz LR PA IS tmp

then simply do as it says and run the command with LR PA IS instead.

I'm sorry that this is not simpler, but having both qform and sform
in the nifti standard makes some things complicated. I am
intending to overhaul the section relating to this and provide
a small script that reorients images to match the MNI152
orientation, as that seems to be what people generally want to do.
It is not important that this is done - it is simply a preference that
people have in order to look at their data. It is only the display
which is now affected by this - all tools process correctly as long
as the qform and sform are consistent and correct. You should
ideally be able to tell if they are correct by looking at the labels
that FSLView attaches to the image. If they are correct and the
qform and sform are consistent then that is enough for all tools
to work. If you also want to see the image in the same orientation
as the MNI152 then perform the fslswapdim commands as shown
above.

I hope this helps to clarify things and make things a little less "blackbox".
Being able to store two different sets of orientation information
and this being somewhat separate from how they are displayed does
make it hard to follow, and I've tried to minimise the amount that
people need to worry about this.

All the best,
Mark

On 14 Jun 2010, at 23:53, Clark Johnson wrote:

Mark -- We use dcm2nii for conversion to NIFTI -- As Chunhui Chen mentioned as this thread began -- this program's output includes an image that has been reoriented -- in my case from a SAG acquisition to what appears in FSLVIEW to be Radiologic.

When viewed in FSLVIEW it "looks" like LAS -- the axial image has the left side of the brain on the right side of the image. But, when you increment the "x" index -- FSLVIEW moves to the right side of the brain. This seems to suggest that FSLVIEW is presenting in LAS -- but the byte order is actually RAS.

If you request the HDR information for this file (fslhd) you get the error

Inconsistent left-right order stored in sform and qform in file XYZ
Using sform instead of qform values

And the output attached to this email.

In your response on May 5th -- you indicated that FEAT / FLIRT won't have any problem with an RAS oriented image (though possibly we have to turn on "Full search" in FEAT). However, the "format and orientation conventions" section of the FSL documentation is still pretty unambiguous about this -- stating:

"NIFTI allows the left-right orientation of your data to be specified but we do not recommend using anything but radiologically ordered images with this release."

It is this statement in the documentation that has me confused. FSLVIEW is pretty clever -- but the byte order in the file does not appear to be LAS -- I am wondering if that means it's not "radiologically ordered".

In a previous response to this issue --

https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1003&L=FSL&D=0&1=FSL&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4&P=179521

you suggested

"If the labels that you see in FSLView (L, R, A, P, S, I) are in
the correct places in the image then all you need to do in order
to reorient your images to axial is:
fslswapdim invol RL PA IS outvol
..."

In an effort to work exclusively with LAS oriented images I've tried the " fslswapdim invol RL PA IS outvol" approach. As noted above, the labels are correct -- I just want the byte order to reflect an LAS orientation as well.

For this file, the suggested fslswapdim call generates several "inconsistent left-right" messages then stops with the following message"

Cannot perform requested swap (NERUOLOGICAL/RADIOLOGICAL storage inverted)
Try the following command instead
Fslswapdim o_dcm2nii_Output.nii.gz LR PA IS tmp

Since I think it already is RAS -- this suggested solution does not seem to solve the problem

We all love FSL tools -- but sometimes they seem a little like a "blackbox" -- things are happening and we don't really understand all of the implications. I am concerned that we are not being careful enough about the L/R orientation of our data and am hoping you can give me some additional guidance

Thank you for all your help.

Clark

-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Mark Jenkinson
Sent: Tuesday, May 04, 2010 3:10 PM
To: [log in to unmask]
Subject: Re: [FSL] Should I flip L-R of images from dcm2niigui (MRIcron)?

Hi,

I'm not quite sure how you are defining RAS, or similar, for fslview
*display*.

Anyway, all that *really* matters is that the labels are in the
correct places on
the image. It doesn't really matter which way the image is
displayed. All the
FSL tools should work, regardless of which way the image is stored and
displayed, as long as the labels are correct. However, if the
different images
are very differently oriented then you should check the registrations
carefully
to make sure that they all work correctly.

So, in summary, I wouldn't worry about swapping anything.

As for FreeSurfer problems, you should email the FreeSurfer list about
that.

All the best,
Mark

On 4 May 2010, at 09:11, chen wrote:

Hello FSLers,

I may be asking a stupid question but it really troubled me a while.

I converted dicom files to nift using dcm2niigui (MRIcron). In
fslview, it displayed in RAS coordinates, but template of FSL are in
LAS coordinates. Should I filp L-R before doing any further analysis?
Besides, my structrual imagews were scanned in sagittal view, and
after dcm2niigui transforation, it's in PIL coordinates in fslview.
Should I reorient the image to LAS coordinates too?
But the letters L R I S A P are labled correctly (at least A-P and I－
S are correct), which means FSL can read in orientation information
correctly. So I am confused whetehr I should reorient my data or not?
I also processed strucutre images with Freesurfer, and the results
between reoriented and not-reoriented images are different, so which
one is correct?

Any help is appreciated,
Thanks!
--
Chunhui Chen

网易为中小企业免费提供企业邮箱（自主域名）

<o_dcm2nii_Output_HD.txt>