Automatic centerline endpoint extraction in Pypes

[Paddy-Xu]

Hi,

I know that the extension in 3D slicer has an Automatic centerline endpoint extraction feature when computing centerline, I am wondering if I there is an option in the script vmtkcenterlines, because my data is really huge and I suspect that running on 3D slicer is less efficient than pure script?

Thanks!

[ramtingh]

This is handled by vmtkcenterlinesnetwork. You'll have to duplicate a lot of what the Slicer script does, as the vmtk version is fairly barebones.

[lassoan]

3D Slicer does not add any overhead to computations. The application is suspended while the algorithm is running, so the CPU is all for VMTK. RAM usage of the application (about 800MB) should be negligible compared to the amount of RAM your computer needs to process large data sets.

The Slicer centerline extraction module is much more advanced than the pypes script for centerline extraction, so someone could consider improving the pypes script. However, interactive features, such as segmentation and centerline extraction are not well suited for batch processing, so I'm not sure if it makes sense to try to implement them using pypes scripts.

[Paddy-Xu]

Hi, thanks for the clarification! But I am considering running it on slurm clusters which means a linux system without user interface, so I might not be able to use Slicer.

[lassoan]

You can use Slicer on a headless machine, for example by using a dummy xserver, such as xdummy.

Vessel segmentation and centerline extraction most often requires some user interaction, which is not well suited for running on a cluster. If your computer cannot process your full data set due limited RAM or CPU then you could downsample it, perform the interactive parts of the workflow on your computer, and then run more demanding processing on the full-resolution data on the cluster. For this you need to customize/combine the scripts in both Slicer and pypes.