smiths
commented
1 year ago I had a quick skim of the linked paper. Great find! You should definitely cite it in your MEng report for future work.
The last figure in the paper (Figure 4) shows a comparison between manual detection of calcification and the algorithm's detection. This is the same idea we have for testing of comparing manual segmentation to algorithm segmentation.
The paper does describe identifying a Volume of Interest in the Methods section in subsection A (Preprocessing).
The insight that the aorta is circular matches the ideas from @inglis-dl. They even talk about oblique transformations to follow the arch, although they don't show the math for this. My guess is that the math is in the paper that they cite (Feuerstein et al).
This article, Automated quantitative 3D analysis of aorta size, morphology, and mural calcification distributions, discussed a fully automated workflow for Aorta Segmentation, however, the steps were very similar to what we have done. It also included the method of finding the aorta center line, so this might be helpful for future work.
What I did not understand is that it did include the crop of volume, which is probably manual work. The segmentation algorithm discussed in the article might be helpful to remove the requirements of user inputs on aorta seeds and other hyperparameters such as LevelSets Segmentation parameters.