Multiple Domain SSM: MLCA pose = shape variation

1

To familiarize myself with multiple domain SSM, I have been following the ShapeWorks use case tutorials (ellipsoid_multiple_domain_mesh and hip_multiple_domain), using the latest version of ShapeWorks (v6.5.1).

Despite following the provided examples with the same settings, my shape analysis results differ from those shown on your website for both ellipsoid_multiple_domain_mesh and hip_multiple_domain.

For the ellipsoid dataset, there are two stacked ellipsoids with identical radii, where the top ellipsoid rotates relative to the bottom one. This represents a pose variation in the data, as the shapes of both ellipsoids remain unchanged. However, when I run the analysis, MLCA identifies the rotating ellipsoid as a shape variation, while almost no pose variation is detected (I sent a video to shapeworks-users@sci.utah.edu “2025-02-27 SW ellipsoid_MD.mp4”). It seems to me that this should be the other way around, as I would expect the variation to be in pose rather than in shape.

In the multiple domain hip dataset, I have ensured that all settings match those in the example on your website (hip_multiple_domain). However, in my results, the hips move apart in a PCA analysis, whereas in the version displayed on your website (previously labeled as “both”), this does not occur (I sent a video to shapeworks-users@sci.utah.edu “2025-02-27 SW hip_MD.mp4”). My assumption is that the use case data may no longer be fully compatible with the latest version of ShapeWorks.

Could you confirm whether these are known issues related to multiple domain SSM? If so, are there any recommended adjustments to ensure correct results?

Thank you in advance for your assistance.

2

I don’t think your videos came through to the list for some reason. You can try emailing me directly at amorris@sci.utah.edu.

I looked into this and have some updates/comments.

  1. I couldn’t actually run the ellipsoid_multiple_domain_mesh use case as is using ShapeWorks 6.5.1. It seems perhaps the dataset changed slightly. I created an issue here:

This has been resolved and merged. I suggest that you try the development build here:

  1. The documentation for these use cases pre-dates the Multi-level PCA analysis, so the screenshots don’t quite match. We need to update them.

  2. There is also an alignment fix in the development version that should make the hip_multiple_domain work properly. We will work on updating the documentation pages.