[Insight-users] DeformableRegistration7.cxx gives worse results

[Emma Ryan]

Hi Luis,

  Thank you for your reply. Please read on for answers to your questions.

------------------>

It shouldn't happen...


           A bit more of information will be helpful...


For example:

1) Are the two synthetic images representing the same
    Modality ?

-----------------> Yes, and intensity distribution is the same. The moving image is generated from the fixed image by rotation and translation.


2) What combination of registration components did
    you used for the Rigid registration ?

      - What Metric ?
      - What Transform ?
      - What Interpolator ?
      - What Optimizer ?

------------------> I use MeanSq metric, Versor3D Transform, Linear Interpolator and regular gradient descent optimizer.
The results after rigid registration are very good. Infact the final  metric value reads 0.75. 


3) Are you feeding DdeformableRegistration7 with the
    resampled image resulting from a Rigid registration
    process ? or did you set up the Bulk transform of
    the BSpline deformable transform ?


------------------> Yes, I use the resampled image resulting from the rigid registration. The setBulkTransform was not available. I see method in the .h file, but when I try to use it, VC++ gives me an error. VC++ could only list the private data member (mbulkTransform) but not the method to access it. So I use the resampled image. But since the moving image is derived from teh fixed, I dont see a problem here.

Also, when I compare the difference images from after performing rigid_registration to before_deformable_registration, the results are not the same. Do you think this is because of resampling as well?  The metric value, obviously is very large (600+), but surprisingly the checkered images in one case were worse after registration.

To give you an example :
   
    One 3D registration took 13 hrs, 20 iterations and reduced the mean squares metric value from 533 to 22. Mean squares metric value after rigid registration was 0.75, the output of which was sent to deformableRegistration7.cpp

Here's the optimizer output:
Starting Registration
Itr  Value       InfinityNormofProjectedGradient
0   533.946   47.9109
1   257.99   37.5374
2   146.285   19.8381
3   105.469   14.2091
4   78.1926   6.82423
5   64.971   5.15543
6   54.1814   3.72431
7   50.4374   5.19165
8   45.1611   2.137
9   41.2881   3.3166
10   36.9809   2.01088
11   34.4687   1.88032
12   31.1311   2.47016
13   29.6365   1.67621
14   27.6934   1.22647
15   26.9191   3.10447
16   25.5062   1.66605
17   24.6931   1.22317
18   23.9171   1.91149
19   23.4049   1.67905
20   22.1302   0.71955


4) How many iterations of the Deformable registration
    did you manage to run before it stopped. ?

----------------->20

BTW, the size limit of emails to the list is 40Kb,
      If you want to show images you could upload
      them to a public web site and send the link
      to the list.

--------------------> I'm sorry, I do not have access to a public site. If I can send you a tar file, I shall try to create one of < 40kb

Thank you for your time and patience,

Emma


Please let us know about the questions above.


   Thanks



     Luis


------------------
Emma Ryan wrote:
> Hi,
> 
>      I am trying to register some synthetic data using 
> deformableRegistration7.cxx from the ITK Examples. The synthetic data is 
> almost well registered using rigid registration. So deformable 
> registration is used to fix the slight mismatch.  Unfortunately, the 
> checkerboard output for the 'after' images are worse that 'before'.
> 
> Why does this happen ?
> 
> Any clues ?  I haven't changed anything in the code except the Pixeltype
> 
> Emma
> 
> p.s. I tried to attach some images but it bounced back.
> 
> ------------------------------------------------------------------------
> Need a vacation? Get great deals to amazing places 
> <http://us.rd.yahoo.com/evt=48256/*http://travel.yahoo.com/;_ylc=X3oDMTFhN2hucjlpBF9TAzk3NDA3NTg5BHBvcwM1BHNlYwNncm91cHMEc2xrA2VtYWlsLW5jbQ-->on 
> Yahoo! Travel.
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users







       
____________________________________________________________________________________
Moody friends. Drama queens. Your life? Nope! - their life, your story. Play Sims Stories at Yahoo! Games.
http://sims.yahoo.com/  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20070726/e40fe723/attachment-0001.htm