Scale in building a statistical shape model

[Maia R.]

Hi,

I have a set of registered meshes and by comparing them, I realized that some meshes do not have the same length: as far as I understand, scale must to be taken into account for the statistical shape model building so do I need to exclude those for the model building ? 

What would be the allowed discrepancy among all meshes to build the shape model ?

Thank you,

Best regards

[Marcel Luethi]

Hi Maia

Whether you want to have scale as part of your model depends on your application. Is the scale likely to be correlated with shape features? If yes, you might want to keep it in the model.

If you want to remove it, simply perform a similarity alignment before model building:

https://github.com/unibas-gravis/scalismo/blob/master/src/main/scala/scalismo/registration/LandmarkRegistration.scala#L105

Best wishes,

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/97c9ca85-2574-470b-bd49-74506afaaa65n%40googlegroups.com.

[Maia R.]

Thank you for the explanation.

For example, for the femur project, if all the surfaces are from adult but some are longer than the majority, is it better I filtered those longer femurs out ? (see on figure, the red one).

In the first course on SSM, you mentionned that a small hand (with small size) must not be included as this could be a baby hand (for anatomical shape).

Thank you very much

Best regards,

Maia

 

[Marcel Luethi]

Hi Maia

It really depends on what you would like to have represented in your model. If scale is of interest to you, because you suspect that there is a morphological change correlated with a change of scale, then you should leave the scale in the model. If you want to analyse the shape only, and are not interested in any information about the size, then you should standardize the size before computing the PCA.

Best regards,

Marcel

To view this discussion on the web visit https://groups.google.com/d/msgid/scalismo/a9f657ee-3783-4dc3-b38b-d95e2b1cce36n%40googlegroups.com.

[V R]

Hi Marcel,

Thank you for your clarification. 

Best regards