[Insight-developers] NIfTI orientation issue

[Bill Lorensen]

Currently (almost 100% of the time) the meta data dictionary is only
populated and accessed in IO classes (I checked this morning). And,
the writers and readers of a given filetype seem to use it correctly.
That is they copy file format specific information from the input to
the output. So I think it is serving a purpose in its current use.

But, as several have noted, it is not propagated through the pipeline.
I think it should be and I'm investigating the solution that Jim
Miller suggested in this bug report:
http://public.kitware.com/Bug/view.php?id=4625

Please stand by,

Bill

On Tue, Oct 27, 2009 at 11:53 AM, kent williams
<norman-k-williams at uiowa.edu> wrote:
> Unfortunately, the MetaData idea has not been consistently implemented, and
> understanding why it's there, and why in particular there is support for it
> in the NIfTI reader/writer means understanding the history.
>
> Hans Johnson did the original addition of Metadata before there was any sort
> of orientation information stored as part of an image. The first reason for
> it being included was to add a way to store orientation as 'out of band'
> data with images.
>
> Now, obviously, having a general metadata infrastructure was valuable from
> the standpoint of, e.g. keeping DICOM tags with an image.
>
> But for the Metadata infrastructure to be consistently implemented, it would
> have to be propogated down pipelines, but it has never been consistently
> implemented.  And it raises issues that aren't easy to definitively settle,
> like what to do when a particular filter makes changes in its output that
> might conflict with the metadata.
>
> Once ITK added orientation information to Images, it made storing
> orientation in the metadata obsolete; and in fact it never was globally
> implemented, so it could never be depended on anyway.
>
> It would be nice if we did a comprehensive review of how metadata is
> actually used, and what it would require to ensure that it's propogation was
> consistent.  But as always, we're all prioritizing available time and
> Metadata keeps getting pushed down.  ITK enhancement and maintainence, at
> least at organizations not directly funded for ITK work, is driven by the
> needs of actually funded research.
>
> On 10/27/09 10:36 AM, "Mathieu Coursolle" <mathieu at rogue-research.com>
> wrote:
>
>> Hi,
>>
>> I am not sure I understand that issue... Is the metadata currently
>> copied in the pipeline ?
>> I also noticed that the metadata in the itkNiftiImageIO has predefined
>> keys starting with ITK_.
>> Is that a standard to follow if I had the metadata from the NIfTI
>> header ? I think that the GDCM reader
>> just uses the DICOM tags as the keys.
>>
>> Thanks,
>>
>> Mathieu
>>
>> On 2009-10-27, at 11:22 AM, Luis Ibanez wrote:
>>
>>> Mathieu,
>>>
>>> No objection from my corner...
>>>
>>> But, please note the related thread regarding the subsequent
>>> passing of the MetaDataDictionary through the pipeline:
>>> http://www.itk.org/pipermail/insight-users/2009-October/033423.html
>>>
>>>
>>>    Regards,
>>>
>>>            Luis
>>>
>>>
>>> -------------------------------------------------------------
>>> On Tue, Oct 27, 2009 at 9:26 AM, Mathieu Coursolle
>>> <mathieu at rogue-research.com> wrote:
>>>> Hi,
>>>> On the same topic, the itkNiftiImageIO has a function called
>>>> DumpNiftiHeader
>>>> used to output
>>>> the header information.
>>>> Any objection on all this header information being added to the
>>>> metadata
>>>> dictionary ?
>>>> Thanks,
>>>> Mathieu
>>>> On 2009-10-26, at 1:59 PM, Mathieu Coursolle wrote:
>>>>
>>>> Hi ITK developers,
>>>> I am using the itkNiftiImageIO to load NIfTI image data. The NIfTI
>>>> file
>>>> format defines 2 different methods to map voxels
>>>> to physical coordinates. One of those methods is using a quaternion
>>>> and is
>>>> defined when the qform_code > 0. The other
>>>> method is using a linear transform and is defined when the
>>>> sform_code > 0.
>>>> Looking at the code which sets the orientation based on the NIfTI
>>>> header, it
>>>> seems to prioritize the qform over the sform.
>>>> (See void NiftiImageIO:: SetImageIOOrientationFromNIfTI(unsigned
>>>> short int
>>>> dims))
>>>> However, there are some cases where I would need both the qform and
>>>> the
>>>> sform transforms once the data is loaded.
>>>> I guess one option would be to add them as part of the metadata ?
>>>> Could we do something like:
>>>> EncapsulateMetaData<std::vector<std::vector<double>
>>>>>> (this->GetMetaDataDictionary(),"qform", myQForm);
>>>> EncapsulateMetaData<std::vector<std::vector<double>
>>>>>> (this->GetMetaDataDictionary(),"sform", mySForm);
>>>> Any other suggestions ?
>>>> Thanks,
>>>> Mathieu
>>>> _________________________
>>>> Mathieu Coursolle, ing.jr., M.Ing.
>>>> Rogue Research Inc.
>>>> www.rogue-research.com
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Powered by www.kitware.com
>>>>
>>>> Visit other Kitware open-source projects at
>>>> http://www.kitware.com/opensource/opensource.html
>>>>
>>>> Kitware offers ITK Training Courses, for more information visit:
>>>> http://kitware.com/products/protraining.html
>>>>
>>>> Please keep messages on-topic and check the ITK FAQ at:
>>>> http://www.itk.org/Wiki/ITK_FAQ
>>>>
>>>> Follow this link to subscribe/unsubscribe:
>>>> http://www.itk.org/mailman/listinfo/insight-developers
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Kitware offers ITK Training Courses, for more information visit:
>> http://kitware.com/products/protraining.html
>>
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-developers
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://kitware.com/products/protraining.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-developers
>