Seeking Guidance on Analyzing Growth-Associated Changes in 3D Models

I am currently working on analyzing surface 3D models at different time points, each corresponding to distinct growth stages. To initiate this analysis, I employed the PseudoLMGeneration method to create pseudo-landmarks on the source model, and subsequently, I transferred all pseudo-landmarks to the target model.

However, given that the surface models exhibit growth in three dimensions, I find myself uncertain about the appropriateness of this method for aligning models of varying sizes at different time points.

As I delve into the intricacies of this analysis, I’m faced with the challenge of exporting the points to GPA for a more in-depth examination of growth-related changes in surface 3D models. If anyone has experience or recommendations in this domain, I would greatly appreciate your insights.

Here’s the steps that I tried to follow.

Thank you so much for your valuable assistance.

Conceptually there is nothing wrong with this workflow. However, there might be practical issues that may cause some problem, particularly when one is working with growth series.

Implicit assumption in this workflow that you can represent your variation with a single template, generated by the PseudoLMGenerators. In my experience that’s a strong assumption for embryonic development, and may not be true. Perhaps maybe at later stages, it might be ok (E18-E19 for mice), but I wouldn’t know without knowing specifics of your data.

The better approach would have been to use stage specific templates to landmark samples, but in that case, you cannot use PseudoLMGenerator. I.e., if you run PseudoLMGenerator with different models, you will get different LM distributions and numbers, hence you can’t do GPA.

Tell me a bit more about your samples (what organism, what developmental stages) so that I can advise accordingly.

Thank you so much for your response. The sample involves the gum model of infants at different time points, and we want to understand how they grow in three dimensions.

In that case, a single template might be sufficient. I would start with a few samples of different age groups and see how well ALPACA registers and transfers the landmarks.

I would probably experiment with a small number manual landmarks first, to find the best parameters for ALPACA. Defaults work for a large number of cases, but it is always worthwhile to try and optimize for your own data.

If the registration and transfer looks good, it is probably ok to use a single template.

How can we assess the good alignment of these geomorphic geometric models if they are all larger in three dimensions?

I would landmark 6-10 manual from my test samples, and use ALPACA, and see how well ALPACA positioned ones align with respect to your manual LMs. Check out the alpaca tutorial for details; specifically this step

Tutorials/ALPACA at main · SlicerMorph/Tutorials (