[Insight-users] MultiResMIRegistration problem

[Luis Ibanez]

Hi Laurent,


1-2) Thanks for letting us know that you managed to read the images.
      Note that VolView *does* create headers for RAW files. Those
      headers are small files with extension .vvi.  If you open a raw
      file VolView will create one of these headers in the same location
      of your .raw file. This header will be used if you read the raw
      image a second time.

3) Rescaling the intensity of the PET image is very important.
    You may want to play with the Intensity Windowing controls on
    VolView (or any other visualization program) in order to find
    a range that will show you anatomical features. This is usually
    in the very low range of PET intensities.  Once you find a range
    you can use the IntensityWindowingImageFilter in order to prepare
    your image.



    The output of the MultiResMIRegistration program (and any other
    ITK registration program for that matter) is the resampled moving
    image using the final transform produced by the registration.
    You *MUST* read the registration chapter from the ITK software guide
    as well as the chapter on image resampling from the same document

               http://www.itk.org/ItkSoftwareGuide.pdf


    You can compare the fixed image and resampled moving image by
    merging them in VolView. Simply use the Merge plugin in the
    Utility group of the Filters menu.




    Regards,


       Luis



-------------------------------------------------
laurent.paul at club-internet.fr wrote:

> 
> 
> Hi luis,
> 
> Thanks for your advices to register my volumes.
> 
> 1-2) I work with volview, so I can crop volumes but the only format I 
> can save for future use is .raw . Volview don't generate a header file 
> for raw datas. That's the reason why I didn't know the*.mha files. Now, 
> I'll edit the *.mha files!
> I use the itkRegionOfInterestImageFilter, and that's great!
> File format isn't a problem anymore.
> 
> 3)I don't know how to rescale intensity. PET images are very dark 
> (pelvis images). Since there's no visible anatomical structures, I'm 
> wondering if it is possible to register that type of volumes. I tried to 
> lighten the volume with shiftfilter plus rescaleintensityfilter but the 
> result isn't very satisfactory. I will try with 
> IntensityWindowingImageFilter.
> I had some results but have still some questions:
> What is the resulting volume of MultiResMIRegistration filter? It seems 
> to be the moving volume on which is applied the rigid transform (and 
> rescale) computed by the filter .
> How can I see the registration result? Is there any filter which merges 
> volumes or overlays slices one by one?
> 
> 4) I read a lot of publications about mutual information to know how it 
> works and know more about filter parameters of MultiResMIRegistrationfilter.
> 
> Thanks,
> 
> Laurent.
> 
> 
> 
> ----Message d'origine----
>  >Date: Sat, 02 Apr 2005 11:15:20 -0500
>  >De: Luis Ibanez
>  >A: laurent.paul at club-internet.fr
>  >Copie à: insight-users at itk.org
>  >Sujet: Re: [Insight-users] MultiResMIRegistration problem
>  >
>  >
>  >
>  >Hi Laurent,
>  >
>  >
>  > Coffee always helps !!
>  >
>  >
>  >Here are some things that you may want to do before passing
>  >your images to the MultiResMIRegistration program:
>  >
>  >1) Convert your DICOM series to a format that uses a single
>  > image. You can do this with the programs in
>  >
>  >
>  > Insight/Examples/IO/
>  > DicomSeriesReadImageWrite.cxx
>  > DicomSeriesReadImageWrite2.cxx
>  >
>  > or with the GUI applicaition MRIConvert
>  >
>  > http://lcni.uoregon.edu/~jolinda/MRIConvert
>  >
>  >
>  >
>  >2) You must crop the CT image in order to contain only the
>  > region that will match the PET image. This will save
>  > *a lot* of computing time, and will prevent the registration
>  > from falling into false positive matchings.
>  >
>  >
>  >3) You should rescale the intensities of the CT and the PET
>  > in order to highlight the structures that you expect to
>  > be matched. For example, body contours. This is important
>  > because Mutual Information uses a joint histogram that covers
>  > the entire dynamic range of each image. If structures that
>  > are supposed to match between the PET and CT images are
>  > restricted to a small region of intensities in the histogram
>  > then the Mutual Information metric will be less sensitive to
>  > the 'goodness' of the matching.
>  >
>  > You can use the filters:
>  >
>  > RescaleIntensityImageFilter
>  > IntensityWindowingImageFilter
>  >
>  > for that purpose.
>  > You want to make sure that after this filters, you images
>  > only have intensities in the range of the anatomical structures
>  > that are visible both in the CT and PET images simultaneously.
>  >
>  >
>  >
>  >4) You should read the documentation of this application:
>  >
>  > InsightApplications/
>  > MultiResMIRegistration/
>  > ReadMe.doc
>  > ReadMe.pdf
>  >
>  >
>  >
>  >5) You can also try the application
>  >
>  > InsightApplications/
>  > LandmarkInitializedMutualInformationRegistration
>  >
>  > (if you don't feel like building it, you can download
>  > the binary version from the UNC CADDLab web site)
>  >
>  > http://caddlab.rad.unc.edu/software/
>  >http://caddlab.rad.unc.edu/software/request.php?file=3DRegistrationTool/3DRegistrationTool_win.zip
>  >
>  >
>  >
>  >Please let us know if you still have any questions,
>  >
>  >
>  > Thanks
>  >
>  >
>  >
>  >
>  > Luis
>  >
>  >
>  >
>  >
>  >-------------------------------------
>  >laurent.paul at club-internet.fr wrote:
>  >
>  >>
>  >> Hi!
>  >>
>  >> I need help! Take a cup of columbian coffee!
>  >> I try to use MultiResMIRegistration example.
>  >>
>  >> To be simple, I have series of Dicom files for CT and Analyze for 
> PET. I
>  >> can convert these in any type of file (except ones I don't know as .mha
>  >> .mhd). I'd want to register these series.
>  >> It would be great if you could say me what type of volume I can use
>  >> (Analyze, serie of 2D images as TIFF, JPG, PNG, Dicom, or other).
>  >> Apparently, the volume is passed by his header file?
>  >> Is the two volumes must have the same dimensions (number of slices) and
>  >> include the same part of the volume? In fact PET embodies the whole
>  >> patient when the CT embodies only the pelvis. Is that a problem?
>  >>
>  >> I saw in users archives that I could use *.mha files. What are that
>  >> files? I have some but can't open them. Tell me more about that format.
>  >> Softwares tell me header is wrong(?!).
>  >>
>  >> I managed to run the application but just for 2D file. The result is
>  >> very bad but that's normal becuse the files I tried to register are
>  >> pelvis images (PET contains few informations, just the tumor). I'd want
>  >> to try before with brain images (less difficult to register I think). I
>  >> saw in the users archives that there are images on the ftp server,
>  >> brainweb165a10f17 and brainweb1e5a10f17Rot10Tx15. I didn't find 
> them. Is
>  >> this application works with 2D images?
>  >>
>  >> Least, I don't understand some parameters.
>  >> Line 5: Multi resolution level? Which level?
>  >> Line 6-7:starting level?
>  >> Line 10: Scale applied?
>  >>
>  >> I know that's a lot of questions but I don't understand Please, just
>  >> give me some keys to register brain images.
>  >>
>  >> Thanks for your help.
>  >>
>  >> Laurent PAUL.
>  >>
>  >>
>  >> 
> ------------------------------------------------------------------------
>  >>
>  >> _______________________________________________
>  >> Insight-users mailing list
>  >> Insight-users at itk.org
>  >> http://www.itk.org/mailman/listinfo/insight-users
>  >
>  >
>  >
>  >