I have created ~100 3D surface models of male and female human skulls from the NMDID. I aim to quantify sexual dimorphism in size and shape:
For size, I have calculated Endocranial Volume (ECV, cm3).
For shape, I want to use a combination of fixed single Lm and semiLM patches at the basicranium.
I need adequate landmarking coverage, but I only want landmarks that quantify variation at the basicranium. I am attempting to use the SemiLMPatch function, and will merge all landmarks (see photos below). The patches are as triangles, and struggles to stay projected at the surface: does anyone have any suggestions to make this easier?
Does anyone have advice to maintain accuracy here, as I do not want to manually landmark each specimen, so ALPACA is being considered, but the landmark alignment seemed slightly ‘off’ during a first run with the template.
Having said that you are trying to landmark a fairly complex area. Any method that involves projecting landmarks will have hard time avoiding those holes and crest etc. Also the landmarks on the basicranium tends to be quite adjacent.
If I were you, I would look into our new extension, Dense Correspondence Analysis, which is designed to tackle exactly situations like this.
See the section on step-by-step landmarking (DeCaL module)
I am assuming I need to manually apply all models with fixed single Landmarks before the DeCaL procedure (creating the atlas etc)?
Thank you so much for the assistance!
You have couple options. The simplest will be use the scissors tool in Dynamic model to crop out the section of the model you want to put landmarks on and then use the grid landmarks. That way you have less complexity to deal with.
If you do not want to do that, you can nudge the four corner points slightly and also play with the projection parameter in the advanced setting.
If you already have manual landmarks for all the models, you are good to go with the DeCaL.
Thank you. The DeCAL method is providing me with pseudo-landmarks for the entire skull model, is this somewhat diluting the shape analysis I specifically wanted at the skull base?
That’s because you did not follow the same LM ordering in all your samples.
To avoid that, I would suggest creating a LM template and then using in all your samples.
Re-landmarking in sequence with the template has worked, and I have managed to generate multiple pseudo-landmarks at the skull base, instead of the entire skull vault:
The final Run subsetting stage appears to be failing however, as I can’t see the new landmarks saved in the output_directory.
I have >100 skulls to landmark before DeCAL, but will use ALPACA and adjust accordingly. Then I will merge the fixed single and pseudo-landmarks together and run GPA/PCA? This has been extremely grateful Murat, thank you!
I can’t replicate this on my end. If you can share your data, I can take a look.
I would not use ALPACA in this context. The main goal of DeCA is to create reliable correspondences, and ALPACA has no constraints on that. You can use ALPACA to transfer manual landmarks to your remaining skulls, but you should review them carefully before you feed them into DeCA.
I am not sure I understand the issue. The subsetted landmarks are saved in the folder you shared. There are 32 LMs in each of them. Isn’t that’s what you wanted?
My apologies, they were in my DeCAL_output file. I now have .json files for both fixed single landmarks (32, which are present) and for the subsetted pseudo-landmarks for each model generated from DeCAL. I will then merge fixed and pseudo-landmarks for each model. Thank you Murat.