[Mrtrix-discussion] csdeconv peaks

Sun Oct 21 17:47:15 PDT 2012

Hi Peter,

Sorry, just noticed you mention your images are LPS, in which case the
primary source of mismatch will probably be the data layout, as suggested
by Rob. The way this was handled for overlays in FSL was to convert all of
the images using mrconvert (from NIfTI straight back to NIfTI), including
the image that you're tryiing to overlay onto. This will modify the
sform/qform as I mentioned previously, and convert all your data to RAS (as
per the NIfTI standard, and MRtrix's internal coordinate system). I have no
idea whether this will fix all the issues (unlikely), but worth a shot...

Cheers,

Donald.

On 22 October 2012 11:37, Donald Tournier <d.tournier at brain.org.au> wrote:

> Hi Peter,
>
> Further to Rob's response, the issue really is that in MRtrix, the
> directions are stored be with respect to real/scanner coordinates
> (technically, the RAS equivalent of the DICOM patient-centered coordinate
> system), not the image axes as (I assume) they are in FSL. The directions
> probably won't match unless the image transform is identity - a straight
> non-oblique axial scan. To make matters worse, as far as I know (someone
> correct me if I'm wrong), FSL's coordinate system is LAS, whereas MRtrix is
> RAS, so even for a straight axial scan the x-axis will probably be flipped.
>
> And then on top of that, there is a good chance that the sform/qform of
> the original NIfTI data corresponds to some other orientation (sagittal /
> coronal), in which case Rob's comments about the data layout would also be
> relevant. Note that MRtrix currently writes out NIfTI images in a
> near-axial layout (the data layout will always be +0,+1,+2,+3), regardless
> of the original qform/sform of the original images, and will modify the
> transform (or the new qform/sform for NIfTI) accordingly. You will only
> notice a mismatch due to the data layout issue if trying to overlay
> MRtrix-derived images/data onto your original NIfTI data (in FSL only,
> MRtrix should handle this appropriately). There was a discussion about the
> overlay in FSL issue a while back on this mailing list.
>
> Also, just to clarify: with MRtrix it's not possible to produce NIfTI
> images with the volume contiguous in RAM (Rob's '-layout +1,+2,+3,+0'
> suggestion), since this is not generally supported by that format
> (technically that's not quite correct, it is possible to specify the data
> type as being a vector, in which case this particular scenario could be
> supported, but since arbitrary strides (layout) are not supported under
> NIfTI, it's not a special case that I thought was worth the trouble to
> implement).
>
> Essentially, this is the inverse problem to that of using bvals/bvecs
> files with MRtrix: you need to convert from scanner coordinates to image
> coordinates to view the directions in FSL...
>
> Not entirely sure this helps you, but at least you should have a clearer
> understanding of why this is happening.
> Cheers,
>
> Donald.
>
>
> On 21 October 2012 12:22, Robert Smith <r.smith at brain.org.au> wrote:
>
>> Peter
>>
>> MRtrix is extremely flexible with regards to how the image data are
>> aligned in memory; if you run the mrinfo command on your image and check
>> the 'data layout' field, you'll most likely see that the ordering of the
>> data in memory in those axes is reversed (i.e. has a '-' sign).
>> Unfortunately FSL doesn't always interpret this in the same way as MRtrix
>> does. There is also some ambiguity in the Nifti file format regarding how
>> the image transform should be stored/interpreted.
>> In my experience, when using MRtrix and FSL in conjunction, the safest
>> option is to run the mrconvert command on your image, with the following
>> option: '-layout +0,+1,+2'. This makes the data layout as unambiguous as
>> possible.
>> In your case with a 4D image, I would try '-layout +1,+2,+3,+0'; this
>> makes the 4th dimension (the fibre orientations) contiguous in memory, so
>> reading the image data for a single voxel is as fast as possible.
>>
>> Hope this fixes your problem; if not let me know.
>> Rob
>>
>> --
>>
>> Robert Smith
>> PhD Candidate
>>
>> The Florey Institute of Neuroscience and Mental Health
>> Melbourne Brain Centre - Austin Campus
>> 245 Burgundy Street
>> Heidelberg Vic 3084
>> Ph: +61 3 9035 7128
>> Fax: +61 3 9035 7301
>> www.florey.edu.au
>>
>>
>>
>>
>> On Fri, Oct 19, 2012 at 9:42 PM, Peter Neher <p.neher at dkfz-heidelberg.de>wrote:
>>
>>>  One more problem. If i apply the peak extraction to my image file the
>>> output nifti peaks seem to have two inverted dimensions. Do you internally
>>> change the image geometry somehow? My image has a left-posterior-superior
>>> orientation.
>>>
>>> Best
>>> Peter
>>>
>>>
>>> On 10/18/2012 10:22 PM, Luis Concha wrote:
>>>
>>> Once you get your peaks image, you can use fslroi to extract each set of
>>> peaks. For example, if you found tree peaks per voxel, the 4th dimension of
>>> your file should have 9 elements (3 peaks, each with xyz components). So,
>>> to get the first peak, you would invoke:
>>>
>>> fslroi my_peaks.nii first_peak 0 3
>>>
>>> and to get the second largest peak:
>>> fslroi my_peaks.nii second_peak 2 3
>>>
>>> and finally the third:
>>> fslroi my_peaks.nii third_peak 6 3
>>>
>>> (I am sure you can substitute fslroi for mrcat for similar operations)
>>>
>>> You can visualize them as lines in fslview.
>>>
>>> Luis
>>>
>>>
>>>
>>> On Thu, Oct 18, 2012 at 3:11 PM, David Raffelt <d.raffelt at brain.org.au>wrote:
>>>
>>>> Hi Peter,
>>>> I would suggest using more directions than SH coefficients in your FOD
>>>> image. Check out the find_SH_peaks help for more information on how the
>>>> peaks are saved. What it does not mention though is that the output peaks
>>>> are ordered by size (largest first). Sorry I'm not familiar with the FSL
>>>> peak file format, however I'm sure it would be fairly trial to code
>>>> a conversion tool up. A good place to start would be dir2amp, since this
>>>> code reads in the directions file and loops over each peak.
>>>> Cheers,
>>>> Dave
>>>>
>>>>
>>>> On 19 October 2012 00:43, Peter Neher <p.neher at dkfz-heidelberg.de>wrote:
>>>>
>>>>>  Hi Dave,
>>>>>
>>>>> seems to work fine. Do you have any recommendation how many directions
>>>>> to generate using gendir? And can you give me the exact specification of
>>>>> the find_SH_peaks output image so that I can read it somehow using my own
>>>>> code? Or is there a conversion to the FSL peak file format?
>>>>>
>>>>> Best,
>>>>> Peter
>>>>>
>>>>>
>>>>>
>>>>> On 10/18/2012 11:37 AM, David Raffelt wrote:
>>>>>
>>>>> Hi Peter,
>>>>> You can use the find_SH_peaks command. Note that you also have to
>>>>> supply a set of directions as input, which are used as seeds for the peak
>>>>> finding. You can generate these using the gendir command.
>>>>>
>>>>>  Note that dir2amp can be used to convert these directions to peak
>>>>> amplitudes.
>>>>> Cheers,
>>>>> Dave
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 18 October 2012 19:06, Peter Neher <p.neher at dkfz-heidelberg.de>wrote:
>>>>>
>>>>>> Hi all, I wondered if it is possible to directly output the fiber
>>>>>> directions and not only the SH representation of the ODFs. If this is not
>>>>>> possible, can anyone tell me how the SH file looks like so that I can read
>>>>>> it and extract the peaks myself?
>>>>>>
>>>>>> Best
>>>>>> Peter
>>>>>> _______________________________________________
>>>>>> Mrtrix-discussion mailing list
>>>>>> Mrtrix-discussion at www.nitrc.org
>>>>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>  --
>>>>> *Dr David Raffelt*
>>>>> Post Doctoral Fellow
>>>>>
>>>>>  The Florey Institute of Neuroscience and Mental Health
>>>>> Melbourne Brain Centre - Austin Campus
>>>>> 245 Burgundy Street
>>>>> Heidelberg Vic 3084
>>>>> Ph: +61 3 9035 7024
>>>>>  www.florey.edu.au
>>>>>
>>>>>
>>>>>   --
>>>>> Dipl.-Inform. Peter Neher
>>>>> German Cancer Research Center
>>>>> (DKFZ, Deutsches Krebsforschungszentrum in der Helmholtz-Gemeinschaft, Stiftung des öffentlichen Rechts)
>>>>> Division of Medical and Biological Informatics
>>>>> Im Neuenheimer Feld 280, D-69120 Heidelberg
>>>>>
>>>>> Phone: 49-(0)6221-42-3552, Fax: 49-(0)6221-42-2345
>>>>> E-Mail: p.neher at dkfz-heidelberg.de, Web: www.dkfz.de
>>>>>
>>>>> The information contained in this message may be confidential and legally protected under applicable law. The message is intended solely for the addressee(s). If you are not the intended recipient, you are hereby notified that any use, forwarding, dissemination, or reproduction of this message is strictly prohibited and may be unlawful. If you are not the intended recipient, please contact the sender by return e-mail and destroy all copies of the original message.
>>>>>
>>>>>
>>>>
>>>>
>>>>  --
>>>> *Dr David Raffelt*
>>>> Post Doctoral Fellow
>>>>
>>>>  The Florey Institute of Neuroscience and Mental Health
>>>> Melbourne Brain Centre - Austin Campus
>>>> 245 Burgundy Street
>>>> Heidelberg Vic 3084
>>>> Ph: +61 3 9035 7024
>>>>  www.florey.edu.au
>>>>
>>>>
>>>> _______________________________________________
>>>> Mrtrix-discussion mailing list
>>>> Mrtrix-discussion at www.nitrc.org
>>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>>
>>>>
>>>
>>> --
>>> Dipl.-Inform. Peter Neher
>>> German Cancer Research Center
>>> (DKFZ, Deutsches Krebsforschungszentrum in der Helmholtz-Gemeinschaft, Stiftung des öffentlichen Rechts)
>>> Division of Medical and Biological Informatics
>>> Im Neuenheimer Feld 280, D-69120 Heidelberg
>>>
>>> Phone: 49-(0)6221-42-3552, Fax: 49-(0)6221-42-2345
>>> E-Mail: p.neher at dkfz-heidelberg.de, Web: www.dkfz.de
>>>
>>> The information contained in this message may be confidential and legally protected under applicable law. The message is intended solely for the addressee(s). If you are not the intended recipient, you are hereby notified that any use, forwarding, dissemination, or reproduction of this message is strictly prohibited and may be unlawful. If you are not the intended recipient, please contact the sender by return e-mail and destroy all copies of the original message.
>>>
>>>
>>> _______________________________________________
>>> Mrtrix-discussion mailing list
>>> Mrtrix-discussion at www.nitrc.org
>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>
>>>
>>
>> _______________________________________________
>> Mrtrix-discussion mailing list
>> Mrtrix-discussion at www.nitrc.org
>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>
>>
>
>
> --
> *Dr Jacques-Donald Tournier
> *
> Research Fellow
>
> The Florey Institute of Neuroscience and Mental Health
> Melbourne Brain Centre - Austin Campus
> 245 Burgundy Street
> Heidelberg  Vic  3084
> Ph:  +61 3 9035 7033
> Fax:  +61 3 9035 7307
> www.florey.edu.au
>
>
>

-- 
*Dr Jacques-Donald Tournier
*
Research Fellow

The Florey Institute of Neuroscience and Mental Health
Melbourne Brain Centre - Austin Campus
245 Burgundy Street
Heidelberg  Vic  3084
Ph:  +61 3 9035 7033
Fax:  +61 3 9035 7307
www.florey.edu.au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20121022/140f5e5e/attachment-0001.html