I am doing a research study with human bones where every bone has a little bit of different area das is marked and I would like to transform a source model exactly to a target bone so I can see the difference between the markings on the target model (.ply). The most important area for the overlap is the joint surface. I tried alpaca but the model ends up in a distorted model and does not look the same. Then I tried the IGT fiducial registration wizard with similarity, which comes very close, but the models are not exactly overlapped. What would be the best way to get a nice overlap of both models? I added 2 models with the markings which I want to overlap.
Are these models the same (e.g., surgical treatment before and after), or two different bones? If latter, you need deformable registration to align them, similarity will make it close, but cannot account for small localized difference.
ALPACA is not a registration tool. Use FastModelAlign (in SlicerMorph), and enable deformable registration
Thank you for your answer. yes these are 2 different bones. I tried Fastmodel along, but I do not find “enable deformable registration” in the module (see attachment). I use slicer 5.10.
Yesterday I tried the IGT first to align with similarity and then run alpaca. Then I got a pretty close match (attached).. The most important overlap has to be the joint surface which looks almost perfect now. However I am not sure if this is really the correct way to do it.
Thanks for the tip, i found it! but the fastmodelalignement with deformation did not the best job, and the new model looks quiet deformed and not identical to the target model.. at the moment the best way was igt similarity followed by alpaca, but i am not sure if this is methodological correct for future scientific publication in a medical journal..
Interesting, FastModelAlign, (including deformable registeration) uses the same library as ALPACA. So it is surprising you get different results. Did you do its rigid registration before hand?
You are right, I did not do rigid registration first.. So now I did a registration with option of scaling and rigid and then I did another registration only with deformable registration and I think the result looks like I had in my mind. I attached the two models and when they are overlapped.
I hope the reviewer will be okay with this technique I will write the methods sth. like that:
“Surface models were registered using the FastModelAlign module in 3D Slicer (v5.11.0 preview). After point-cloud subsampling, global scaling and rigid alignment were performed. Non-rigid registration was subsequently applied using a CPD-based deformation framework, and the resulting deformation field was interpolated onto the original surface mesh for downstream analyses.”
I will write the methods sth. like that:
