Dear Robin,
There does seem to be a problem in Tiger which doesn't occur
anywhere else. However, it only occurs when the old-style syntax
in convert_xfm is used (where images are specified) whereas the
new-style syntax of just using matrices works fine. So I've removed
all the old-style syntax from updatefeatreg (-in, -ref and -middlevol)
and it seems to work. I'm attaching a copy of the edited script
here, so try putting this script in $FSLDIR/bin and then running
it in your feat directory as normal.
I think this should work fine - but let me know if you have any
problems with it.
All the best,
Mark
On 18 Oct 2005, at 23:34, Robin Goldman wrote:
> Hi Mark,
>
> I had copied all the images into the directory to make sure links
> were not the problem. All the images look fine in fslview.
>
> Do you want the whole feat directory, or just the reg directory?
>
> Thanks,
> R
> ________________________
> Robin Goldman, Ph.D.
> Hatch Center for MR Research
> Columbia University
> 710 W. 168th Street, NIB-1
> New York, NY 10032
> (212) 342-0867
>
> On Oct 18, 2005, at 5:53 PM, Mark Jenkinson wrote:
>
>
>> Dear Robin,
>>
>> Convert_xfm shouldn't care whether it uses analyze or nifti
>> images. The errors that you are seeing here are consistent
>> with what happens when an image is corrupted or does not
>> exist (which could happen if the link was broken).
>>
>> Can you load the highres image here into FSLView and does
>> it look OK?
>>
>> Try copying the image files into this directory too if they
>> happen to be links at the moment. That might fix it,
>> especially if they are not standard unix links, but the
>> strange mac aliases instead.
>>
>> Finally, can you try executing that single convert_xfm call
>> from the command line and see if it works or gives a
>> different error message.
>>
>> If all that still fails to get the desired results then
>> I can't think of anything else I can do remotely.
>> In this case, please tar up the directory (but make sure all
>> images are copies, not links) then upload the tar file to us at:
>> http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
>> and email me directly with the ID number for the upload session.
>>
>> Hopefully we can get to the bottom of this soon.
>> All the best,
>> Mark
>>
>>
>> On 18 Oct 2005, at 20:44, Robin Goldman wrote:
>>
>>
>>> is convert_xfm now not compatible with analyze images? Does it
>>> only read nifti?
>>>
>>> [ni0bgoldman:R21_OddBall/RG_081805/
>>> volunteerrobin_18082005_epi_6823.feat] rig% updatefeatreg .
>>> ** ERROR: nifti_image_read(highres): can't open header file
>>> ** ERROR: nifti_image_open(highres): bad header info
>>> Error: failed to open file highres
>>> ERROR: Could not open image highres
>>> Image Exception : #22 :: Failed to read volume highres
>>> /usr/local/bin/fsl/bin/updatefeatreg: line 90: 843 Abort
>>> trap ${FSLDIR}/bin/convert_xfm -in example_func -ref
>>> standard -middlevol highres -omat example_func2standard.mat -
>>> concat highres2standard.mat example_func2highres.mat
>>> ** ERROR: nifti_image_read(highres): can't open header file
>>> ** ERROR: nifti_image_open(highres): bad header info
>>> Error: failed to open file highres
>>> ERROR: Could not open image highres
>>> Image Exception : #22 :: Failed to read volume highres
>>> /usr/local/bin/fsl/bin/updatefeatreg: line 95: 844 Abort
>>> trap ${FSLDIR}/bin/convert_xfm -in highres -ref
>>> standard -inverse -omat standard2highres.mat highres2standard.mat
>>> ** ERROR: nifti_image_read(highres): can't open header file
>>> ** ERROR: nifti_image_open(highres): bad header info
>>> Error: failed to open file highres
>>> ERROR: Could not open image highres
>>> Image Exception : #22 :: Failed to read volume highres
>>> /usr/local/bin/fsl/bin/updatefeatreg: line 98: 845 Abort
>>> trap ${FSLDIR}/bin/convert_xfm -in example_func -ref
>>> highres -inverse -omat highres2example_func.mat
>>> example_func2highres.mat
>>>
>>> [ni0bgoldman:R21_OddBall/RG_081805/
>>> volunteerrobin_18082005_epi_6823.feat] rig% ls reg/
>>> example_func.nii.gz@ example_func2standard.xfm
>>> highres2standard.xfm
>>> example_func2highres.gif highres.hdr@
>>> index.html
>>> example_func2highres.mat highres.img@
>>> standard.hdr@
>>> example_func2highres.xfm highres2example_func.mat
>>> standard.img@
>>> example_func2standard.gif highres2standard.gif
>>> standard2example_func.mat
>>> example_func2standard.mat highres2standard.mat
>>> standard2highres.mat
>>> [ni0bgoldman:R21_OddBall/RG_081805/
>>> volunteerrobin_18082005_epi_6823.feat] rig%
>>>
>>> ________________________
>>> Robin Goldman, Ph.D.
>>> Hatch Center for MR Research
>>> Columbia University
>>> 710 W. 168th Street, NIB-1
>>> New York, NY 10032
>>> (212) 342-0867
>>>
>>> On Oct 18, 2005, at 12:04 PM, Robin Goldman wrote:
>>>
>>>
>>>> Hi Mark,
>>>>
>>>> Thanks. I ran updatefeatreg again, and still get a segmentation
>>>> fault. I'm testing everything in feat directories that were run
>>>> in full, so I know the reg directories contain everything they
>>>> should (and I double checked as well). Running on OS X 10.3.9,
>>>> here's what I get --
>>>>
>>>> [ni0bgoldman:R21_OddBall/RS_092105/
>>>> rs_r21odd_21092005_epi_tr_sync1_7837_brain.feat] rig%
>>>> updatefeatreg .
>>>> /usr/local/bin/fsl/bin/updatefeatreg: line 98: 796
>>>> Segmentation fault ${FSLDIR}/bin/convert_xfm -in
>>>> example_func -ref highres -inverse -omat
>>>> highres2example_func.mat example_func2highres.mat
>>>>
>>>> What happens when this script gets a segmentation fault at this
>>>> point? It seems that it is the inverse matricies that cause the
>>>> problem. Will the other essential files for higher level stats
>>>> be created anyway? The inverse matrices aren't necessary for
>>>> higher levels, are they?
>>>>
>>>> Best,
>>>> Robin
>>>> ________________________
>>>> Robin Goldman, Ph.D.
>>>> Hatch Center for MR Research
>>>> Columbia University
>>>> 710 W. 168th Street, NIB-1
>>>> New York, NY 10032
>>>> (212) 342-0867
>>>>
>>>> On Oct 16, 2005, at 9:35 AM, Mark Jenkinson wrote:
>>>>
>>>>
>>>>> Hi Robin,
>>>>>
>>>>> The essential files required by feat in the reg directory are:
>>>>>
>>>>> example_func
>>>>> highres
>>>>> standard
>>>>>
>>>>> example_func2highres.mat
>>>>> highres2standard.mat
>>>>>
>>>>> example_func2standard.mat
>>>>> highres2example_func.mat
>>>>> standard2highres.mat
>>>>> standard2example_func.mat
>>>>>
>>>>>
>>>>> The first three are images, or links to images, and you should
>>>>> be able to view them with fslview. If they view OK then that
>>>>> should be fine.
>>>>>
>>>>> The next two files are the affine transformation files that
>>>>> are created by running flirt to register the pairs of images
>>>>> together (first the example_func to the highres, then the
>>>>> highres to the standard). These are the most essential two
>>>>> matrix files, as the others are derived from these two, and
>>>>> updatefeatreg *should* generate all other matrix files, but
>>>>> relies on these being present.
>>>>>
>>>>> Finally, the other affine matrix files are either concatenated
>>>>> versions of the above two affine matrices (in the case of
>>>>> example_func2standard.mat) or are inverses of the other
>>>>> matrices, including the concatenated one.
>>>>>
>>>>> So if you need to do everything by hand, then you can
>>>>> populate the reg directory with all of these files. However,
>>>>> if you have the three images and the essential two matrix
>>>>> files then updatefeatreg should work OK. Try running
>>>>> updatefeatreg from the feat directory itself. That is:
>>>>> cd blah/blah/DY004_novOdd_0_TestRegOnly.feat
>>>>> updatefeatreg .
>>>>> Hopefully this will work better, as sometimes directory
>>>>> links or mountings can make relative paths (like you
>>>>> used in your example with ../../ a little tricky and
>>>>> error prone).
>>>>>
>>>>> So I would try and get updatefeatreg to work, as it is
>>>>> designed to do exactly what you want. However, in
>>>>> your case you must make sure that you link (or copy) the
>>>>> standard image into the reg directory as well as creating
>>>>> a highres2standard.mat file by running flirt to register
>>>>> the highres to the standard image. Once these are in
>>>>> place (along with the example_func2highres.mat) then
>>>>> updatefeatreg should be fine. Let me know if it isn't.
>>>>>
>>>>> All the best,
>>>>> Mark
>>>>>
>>>>>
>>>>>
>>>>> On 14 Oct 2005, at 23:28, Robin Goldman wrote:
>>>>>
>>>>>
>>>>>> Hi there,
>>>>>>
>>>>>> I have run feat on a dataset without registering the images to
>>>>>> standard space, and now I want to
>>>>>> add the srandard registration to the reg directory. I have
>>>>>> looked at the FAQ, but since I did not run
>>>>>> the registration with the standard image originally and for
>>>>>> some reason updatefeatreg is crashing
>>>>>> on me (see my prior post), I'm looking for a work around. If I
>>>>>> add a link to the standard image and
>>>>>> a highres2standard.mat, what else is needed to run the higher
>>>>>> level stats? I don't need to have it
>>>>>> all linked to the report page...
>>>>>>
>>>>>> Best,
>>>>>> Robin
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>
>
|