GroupWise: parametric, non-rigid registration

[Bereket Workie Anteneh]

Dear Scalismo community,

I am doing parametric, non-rigid registration for the SSM of proximal femur using the code provided in lecture  12 (https://scalismo.org/docs/tutorials/tutorial12).  For SSM building I have 28 number of datasets, from these datasets I use one of them as a reference mesh (say ''set1'' ) and the remaining datasets (27) as my target datasets. I want to do a GroupWise registration (that mean, I want to align all the 27 datasets to the reference mesh simultaneously. In the lecture 12 of the code the registration is declared as Pairwise registration ( that mean, individual datasets are aligned with the  reference mesh separately, i.e. first, reference with dataset 2, then reference with dataset 3,....., dataset 28.

I want to do a GroupWise registration (that mean, I want to align all the 27 datasets with the reference mesh simultaneously (once). 

So, how I can declare the code to do GroupWise: parametric, non-rigid registration?

With regards,

[Marcel Luethi]

Dear Bereket

Group-wise registration is not something that Scalismo currently offers. It is something you would need to build on top of the available building blocks provided by Scalismo.

What is your use case? Which benefits do you hope to get from group-wise registration?

Best regards,

Marcel

--
You received this message because you are subscribed to the Google Groups "scalismo" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scalismo+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/scalismo/883db1e7-8ad3-47e5-83fe-d4b440720cd6n%40googlegroups.com.

[Bereket Workie Anteneh]

My intention to use GroupWise registration methods was to mitigate uncertainties associated with any one mesh by simultaneously registering all meshes in a population. 

This incorporates all mesh information in registration process and eliminates bias towards a chosen reference frame.

Anyways, if currently Scalismo does not offer GroupWise registration, as you said  ''building on top of the available building blocks provided by Scalismo'' might be mandatory. 

My question here is, how I would build on top of the available building blocks provided on the Scalismo?

With regards,

[Marcel Luethi]

My advice would be to start with pairwise registration, identify the concrete problems that you experience and plan the next steps from there.

The building blocks are GPs, the current registration framework (including metrics, regularizers) as well as the full sampling framework. These can be used or extended as ready made parts in a project that implements groupwise registration.

Best regards,

Marcel