[Insight-users] Deformable Registration

Cristina Ciavarro cristinaciavarro at hotmail.com
Mon May 26 08:25:35 EDT 2008


Dear itk users, 
I have a question about the deformable registration.

I'm using DeformableRegistration8 for the registration between a reference image (512x512x83 pixels; resolution: 0.976562x0.976562x2.5 mm3) and a test image (552x552x96; resolution: 0.976562x0.976562x2.5 mm3).

I set:
grid = 17x17x7 (+3);
number of multiresolution level = 3;
optimizer:
       - first level
       optimizer->SetCostFunctionConvergenceFactor( 1e+12 );
  	optimizer->SetProjectedGradientTolerance( 1e-4 );
            -second level
	optimizer->SetCostFunctionConvergenceFactor( 1e+7 );
  	optimizer->SetProjectedGradientTolerance( 1e-6 );
            - third level
	optimizer->SetCostFunctionConvergenceFactor( 1e+7 );
  	optimizer->SetProjectedGradientTolerance( 1e-8 );

GLOBALLY,I think that this strategy was good: infact there is a return of the lung volume, compared with the only rigid registration. But there is a local problem. I gave the INPUT POINTS and I get the new coordintes by TrasformPoint: there is an error that often doesn't reduce compared with the error got by the rigid registration. I also trayed with 2 steps of deformable registration, setting:
grid = 17x17x7 (+3);
number of multiresolution level = 3;
optimizer:
       - first level
       optimizer->SetCostFunctionConvergenceFactor( 1e+12 );
  	optimizer->SetProjectedGradientTolerance( 1e-4 );
            -second level
	optimizer->SetCostFunctionConvergenceFactor( 1e+7 );
  	optimizer->SetProjectedGradientTolerance( 1e-6 );
            - third level
	optimizer->SetCostFunctionConvergenceFactor( 1e+1 );
  	optimizer->SetProjectedGradientTolerance( 1e-10 );
So, I made twice the deformable registration: in the second time I used the output of the first step as the test image. But there isn't an improvement.

I ask you if there is a way to improve the method, if there is the possibility to operate locally.

Thank you!
Best regards, 
Cristina
_________________________________________________________________
Crea il tuo blog su Spaces, condividi le tue esperienze con il mondo! 
http://home.services.spaces.live.com/


More information about the Insight-users mailing list