I am currently working on a dataset of left ventricles. However, I am having some issues with the point correspondence, specifically at the ventricle’s base.
I’m attaching screenshots of the original and reconstructed ventricles, as well as the optimization parameters I’m using at the moment. I’m working on a larger dataset but using 23 shapes at the moment and 128 particles for the optimization step before extending it.
Does anyone know what I should change to achieve better correspondence and specificity? Or any suggestions on how to adjust the parameters for ventricular shapes?
Thanks in advance for the help!
Do the default parameters look any better? Specifically, you might try disabling the “multiscale mode” and use the default relative weighting options.
Also, if you could post a picture of the mean shape, we might more easily be able to see how good/poor the correspondence is.
Thank your reply and sorry for my late reply, the default parameters indeed worked better and optimizing the shape as a closed surface worked better than with the plane constraint.
It turned out I didn’t fully understand how to check whether the correspondence is achieved, I understood it better when comparing the reconstructed meshes with the original meshes through the surface-to-surface distance and it turned out to work quite well with the default parameters.
Thank you for your help!