Registration mask and rigidity penalty

[Krishna Nanda]

Dear Elastix team,

I am using registration masks both on the fixed and moving images to focus the registration on the regions of interest. So, the image samples are drawn within the fixed image mask for computing the derivative of the similarity as well as bending energy terms during the stochastic gradient descent procedure. I want to also use a rigidity penalty term, with the corresponding rigidity co-efficient image specified on the moving image. Now, my moving image registration mask excludes the region that needs to remain rigid on the moving image during the registration (specified via the rigidity co-efficient image). In this case, does the rigidity penalty term have any effect on the registration?

In the rigidity penalty paper (Staring et al. 2007, Medical Physics), I see that the rigidity term is evaluated on the control points of the B-spline grid and not on the samples drawn from the fixed image inside the fixed image registration mask. In this case, can it happen that even if the fixed image samples do not fall inside the specified rigidity region on the moving image, a control point may, and hence rigidity of the moving image region is imposed?

Thank you!

[Krishna Nanda]

Any advice on this would be greatly helpful. Thank you!

[m.st...@lumc.nl]

Dear Krishna,

If I recall correctly, you specify a separate mask in the parameter file to specify the region that should remain rigid. I am not sure what happens if you excluded this region with your normal mask, but you can inspect the source code to know this for sure:

https://github.com/SuperElastix/elastix/tree/develop/Components/Metrics/RigidityPenalty

The options DilateRigidityImages and DilationRadiusMultiplier manipulate the influence of a region around a control point. In principle it is not possible to keep a single control point rigid, but you need to look at the surrounding control points as well. These parameters enable that.

HTH, Marius

---

From: elastix-imag...@googlegroups.com <elastix-imag...@googlegroups.com> on behalf of Krishna Nanda <krishn...@gmail.com>
Sent: 18 February 2022 1:07 AM
To: elastix-imageregistration
Subject: [elastix-imageregistration] Registration mask and rigidity penalty

 

--
You received this message because you are subscribed to the Google Groups "elastix-imageregistration" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elastix-imageregis...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elastix-imageregistration/54da6715-b3ee-43c0-88a5-ba90e6129983n%40googlegroups.com.

[Krishna Nanda]

Thanks for your reply prof. Staring. 

I looked at "itkTransformRigidityPenaltyTerm.hxx" and could follow at a high level what is going on. But I was not successful in locating the interplay between the registration mask and rigidity co-efficient image (rigidity mask). Intuitively, the sampled points used for computing the registration similarity will not fall inside the rigidity mask region since it is outside the moving image registration mask. So, this should not affect the registration. But if the rigidity penalty is imposed via the B-spline control points, then I can imaging the control point falling inside the rigidity images if not the sampled image points, and hence rigidity mask having an effect. Does this sound correct to you? Any advice would be helpful. Thank you!