Hiya! I’m currently trying to apply sliding semi-landmarks to a dataset (x20 crania/scapulae/clavicles/pelvises) with fixed landmarks already applied. I tried to use the curves to create SSL between landmarks as per Rolfe et al., 2021, however this approach took a lot of time and would slow down the application, therefore I deemed it inefficient. Is there another approach (e.g. using PseudoLM) that would allow me to create the same number of SSL per structure in the same position (e.g. mid-point between landmarks A and B etc…) and replicate this using ALPACA? Thanks 
The answer depends on area you are trying to landmark as well as your landmark distribution. Without seeing that it is hard for me to suggest something, but you might want to give the recent grid landmarks a try.
Hiya Murat,
That’s excellent, I will give this a go on my structures!
Are there any modifications I need to make to the .json file if I’m exporting these sliding semi-landmarks, e.g. saving as ‘description:semi’ when merging, changes to R code when exporting etc…?
I’ve attached images of my semilandmarks so far - they appear to be more uniformly distributed than the PseudoLM generated LMs and allow me to capture specific areas (e.g. the birth canal), however I’m concerned by the time taken to complete each one as well as the relatively low number of SLMs compared to automated methods - is this sustainable?
Many Thanks,
Eve
This is something only you can decide. PseudoLMGenerator is fast, but cannot generate same number of landmark on every sample. So you will have to use a method to transfer those landmarks to individual subjects. You can do that via ALPACA (which uses a point-cloud registration) or projectSemiLMs (which uses a landmarks to transfer).
The other alternative is a new extension we introduced: DenseCorrespondenceAnalysis (DeCA). You will need some fixed landmarks to start the process, but then it will both build a template and then transfer semi-landmarks to the subjects. See GitHub - SlicerMorph/SlicerDenseCorrespondenceAnalysis: Dense Correspondence Analysis for Surfaces (and the tutorial linked there).
ps. Couple issues.
- You are not creating sliding landmark. You are generating semi-landmarks. Sliding is a technique used during the superimposition of landmarks, and specifically applied to the landmarks designated as “semi or pseudoLMs” (io.e., they are not fixed)
- You should not analyze landmarks that belong to different anatomical systems jointly. A major assumption in the analysis is that landmarks designate a single rigid body. Cranium and mandible together is not a single rigid body (they rotate with respect to each other through TMJ).