[Insight-users] Re: ITK Questions
Stephen R. Aylward
aylward@unc.edu
Fri, 15 Nov 2002 18:17:11 -0500
Hi Ofri,
There is documentation on MetaIO in Insight/Utilities/MetaIO/doc
About writing Meta headers - instead of using the example program, you
can also do it using any text editor.
About meta-info regarding hospital and all, you can derive the
MetaImageIO class to include additional fields, you'll just need to add
the tags to the SetupReadFields and SetupWriteFields member functions in
the derived class. Nice thing is that then you'll still be able to
read/process those images using any app that can read metaImages - those
apps will automatically just skip the new/unknown fields.
About reading dicom...you can read most dicom objects using our dicomIO
library. The hard part is creating a 3D block of data from a dicom
object...we hope to have an answer to this soon...can any user recommend
a PD package that does this automatically for GE and Siemens and other
dicom files????? We're hunting for one...
Thanks,
Stephen
Ofri Sadowsky wrote:
> 1. OK. Are the test programs updated too?
>
> 2. I looked at MetaImageImporter and got the idea of how to use it. But it
> would be more helpful if the MetaImageIO format were documented in detail
> somewhere. I don't want to rely on the manual process in MetaImageImporter
> only. Then, I was not sure how to use MetaImageIO. In particular, MetaImageIO
> defines the function
> MetaImage * GetMetaImagePointer (void)
> but no link to class MetaImage is provided, and I could find no reference for
> it in the manual.
>
> 3. By meta-information I mean items such as the ones in Dicom: hospital name,
> patient age, x-ray energy, etc., which may vary between images and according
> to the needs of the user. Neither ImageIOBase nor MetaImageIO has any methods
> to handle that. I think it's hard to give a general format to
> meta-information, but it may prove useful when I want to customize and label
> images. For example, it can provide time stamps for image sequence (video
> etc.), bed position, or any other parameter by which the images are
> parametrized. Maybe the MetaImage class contains something like that and
> maybe not. Maybe the simplest answer is to write my own format and reader.
> But at least for the dicom format, it would be convenient if ITK provided
> methods to extract meta-information, e.g. by tag code.
>
> 4. I'm sending this email to the list as well. I hope that's fine.
>
> 5. My installed version of ITK is 1.0.0. I could start a CVS tracking of
> changes, but it has some drawbacks. For once, I want to use a more or less
> stabilized interface. Then, I don't want to rebuild the ITK libraries every
> other day, or even every week. Do you have a recommended update strategy for
> "external" users? How stabilized are the release interfaces?
>
> Thanks,
>
> Ofri.
>
> Luis Ibanez wrote:
>
>
>>Hi Ofri,
>>
>>Please find below comments/answers to your questions.
>>
>>1) You are right about the inconsistency on the DicomImageIO
>> object. The constructor and destructor should be protected.
>> I just fixed this in the repository. The DicomImageIO object
>> is intended to be used a any other ImageIO object.
>>
>> Something like:
>>
>> itk::DicomImageIO::Pointer dicomIO =
>> itk::DicomImageIO::New();
>>
>> typedef itk::ImageFileWriter<ImageType> WriterType;
>> WriterType::Pointer writer = WriterType::New();
>>
>> writer->SetImageIO( dicomIO );
>>
>>2) The easiest way at this point to read a group
>> of dicom slices into a volume is to write a
>> MetaImage header for them. This is pretty simple
>> to do. The application: MetaImageImporter will
>> guide you step by step through the process.
>>
>> You basically create a text header that lists the
>> names of the dicom files as well as the basic
>> information about spacing and size.
>>
>>3) Well, this is already what you get from the MetaImage
>> reader. Have you tried this approach ?
>> Do you see any serious drawbacks on it ?
>> BTW: ITK is not a Kitware only product. It is
>> developed by a consortium integrating many different
>> groups. Kitware is however strongly commited to
>> maintain and further develop the toolkit.
>>
>>4) Sure,
>> here is the link to the web site where you can
>> suscribe. This is mailman, it is pretty flexible
>> as to how you want to customize your suscription
>> to the list.
>>
>> http://www.itk.org/mailman/listinfo/insight-users
>>
>>Please let me know if you have any other questions.
>>
>>Thanks
>>
>> Luis
>>
>>-------------
>>
>>Ofri Sadowsky wrote:
>>
>>
>>>Hi Luis,
>>>
>>>I have some more questions regarding itk::DicomImageIO and related
>>>classes.
>>>
>>>1. I was looking at the test code provided with ITK at
>>>itkDicomImageIOTest.cxx . In line 43 it calls
>>>itk::DicomImageIOFactory::RegisterOneFactory();
>>>Then, in the doxygen on the web site, the way to use itkImageIOBase is
>>>to plug it into an itkImageFileReader. That is consistent with the
>>>example in itkPNGImageIOTest.cxx.
>>>Also, in itkPNGImageIO the ctor is protected, and an object is created
>>>by the New() factory method. But in the doxygen of itkDicomImageIO the
>>>ctor is public, which is inconsistent again.
>>>So what is the right way to use itkDicomImageIO?
>>>
>>>I ran the IO test on one image that I have, and it seems to run without
>>>problems. At least that means that it works.
>>>
>>>2. We had a previous discussion about the difficulties of processing
>>>DICOM files. Since there is a problem in reading a set of slices and
>>>ordering them correctly, especially when the inter-slice spacing is
>>>nonuniform, I thought I could do at the beginning with processing each
>>>slice separately. The itk::DicomImageIO provides the public methods for
>>>setting a file prefix or a file pattern, but I could not understand how
>>>the range of numbers in the pattern can be provided to the DicomImageIO.
>>>Also, I should be able to find the bed position from the dicom file (the
>>>third coordinate). Is there any method in ITK for that? I guess that you
>>>can't associate this number with ITK 2D image, but at least I'd like to
>>>have some access to it.
>>>
>>>3. In the long run, it might be useful to have some Dicom tag-set that
>>>stores all the meta information on a study (or an image) in one
>>>structure. Are there any plans in Kitware to create such a structure?
>>>
>>>4. I think there was a mailing list of ITK developers/users that you
>>>refered me to previously, but I lost the email. Could you send it to me
>>>again please?
>>>
>>>Thanks,
>>>
>>>Ofri Sadowsky
>>>
>>>
>>>
>>
>
> _______________________________________________
> Insight-users mailing list
> Insight-users@public.kitware.com
> http://public.kitware.com/mailman/listinfo/insight-users
--
===============================================
Dr. Stephen R. Aylward
Assistant Professor of Radiology
Adjunct Assistant Professor of Computer Science
http://caddlab.rad.unc.edu
aylward@unc.edu
(919) 966-9695