Hi,

I ran 10 cases of pulmonary trunk through the SPHARM-PDM pipeline, setting iteration to 2000 (because for the pulmonary trunk cases, the _para.vtk file does not get generated properly from GenParaMesh when using a lower iteration number). 2 out of the 10 cases were not able to output the _para.vtk file.

I checked the two cases's post processed files, and there were no enclosed holes in the segmentations. Also the topology for all 10 cases are quite similar. I re-ran the two cases with iteration set to 2000, and I noticed that the GenParaMesh process had an early termination (within the first 500th iteration) because of NAN coordinates. How is that possible?

What is causing the early termination? Would there be something I can do to output the _para.vtk files properly for these two cases?

Thanks,
Sylvia

Hi Sylvia,


It seems for your description you are handling the problem in the right way. In my experience when a NaN problem ocurrs is usually related to a "problematic topology". This means that even though there are no holes that could make the topology non-spherical, there are very complicated items in the topology that make the program to get to a numerical error and fail. Some of them can be: C-shape topology (that looks like an almost hole), very dented areas, very thin areas, etc etc. I would revise the cases in which you get to the numerical error and try to modify the segmentation searching for the things I just described.

I hope that helps!
Regards,


BEATRIZ