[Insight-users] Exceptions with Deformable BSpline registration.

[johnjtchen at aol.co.uk]

Hi Luis,

Thank you.

It seems that the image orientation of my data is the reason 
that cause this error.

The Print() dumps:
? Direction: 
1 0 0
0 -1 0
0 0 1

After I force the line 245 In DeformableRegistration13.cxx to 
[ 1 0 0
? 0 1 0
? 0 0 1] with 

FixedImageType::DirectionType gridDirection = fixedImage->GetDirection();
gridDirection.SetIdentity();

The registration of example 12, 13,?15 runs quite well!

>>Do you get this exception error at the very beggining of the?
>>registration process ??

I get the error message right after registration->StartRegistration();
It didn't even reaching the optimization stage.


I am wondering that with the same image direction in both fix and
mov images with 

? Direction: 
1 0 0
0 -1 0
0 0 1,

why the registration wouldn't be able to matching two the same 
image direction together? 

Has it got trouble with getting derivatives within BSplineTransform?


Thanks very much.
john




-----Original Message-----
From: Luis Ibanez <luis.ibanez at kitware.com>
To: johnjtchen at aol.co.uk
Cc: insight-users at itk.org
Sent: Sun, 29 Jun 2008 1:20 am
Subject: Re: [Insight-users] Exceptions with Deformable BSpline registration.


Hi John,?
?
This error message indicates that under the current Transform,?
the two images don't have a a significant overlapping region.?
?
This may be the result of poor initialization, or the result?
of an optimizer set to take strides that are too large.?
?
Setting the CMake flag to?
?
? ITK_USE_ORIENTED_IMAGE_DIRECTION=ON?
?
is the right thing to do.?
?
Do you get this exception error at the very beggining of the?
registration process ??
?
or do you get the error message after running the optimzer?
for several iterations ??
?
Could you please do the following ??
?
In any of the examples that you are trying, please add the?
following code:?
?
? fixedImageReader->Update();?
? movingImageReeader->Update();?
? fixedImageReader->GetOutput()->Print();?
? movingImageReader->GetOutput()->Print();?
?
This will help us see what is the initial overlap?
between the images. Please post to the list the entire?
output. We will be particularly interested in Origin,?
Spacing and Direction of both the Moving and Fixed images.?
?
?
Also,?
?
Could you please post to the list the output log of?
DeformableRegistration15 ??
?
This is the example that perfoms Rigid, then Affine, and?
then a BSpline deformable registration with two levels?
of resolution in the BSpline grid.?
?
? Thanks?
?
? Luis?
?
---------------------------?
johnjtchen at aol.co.uk wrote:?
> Dear All,?
> > I have some problems with running BSpline with Mattes MI metric?
> registration from example\registration directory.?
> > No matter how I try the various conbination of the paramters of?
> LBFGSB or RegularStepGradient optimizer, the results that I can?
> get are "Joint PDF summed to zero" with gradient optimizer or?
> "Too many samples are mapping outside image" with LBFGSB.?
> > The images for registration are T1 256*256*92 brains. I am sure?
> that the origin of fix and mov images are both [0,0,0], and the?
> images have been rigidly aligned.?
> > I am windering is that the effects with regarding to the CMake?
> option "ITK_USE_ORIENTED_IMAGE_DIRECTION=ON"? Because?
> now the itkMattesMutualInformationImageMetric has been redirect?
> to itkOptMattesMutualInformationImageMetric.?
> > The examples that I try were Deformable8, 12, 13 and 15. It all gave?
> me the exceptions that mentioned above.?
> > Could you give some advice on this problem??
> > Thanks very much.?
> john?
> > _______________________________________________?
> Insight-users mailing list?
> Insight-users at itk.org?
> http://www.itk.org/mailman/listinfo/insight-users?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20080629/13873faf/attachment-0001.htm>