Lung CT Segmenter supports TotalSegmentator nnU-Net deep learning

A new engine has been added to “experimental AI” in the Lung CT Segmenter module of the Lung CT Analyzer extension: TotalSegmentator by Jakob Wasserthal (University of Basel and colleagues from DKFZ). Please cite the authors.
TotalSegmentator involves nnU-Net deep learning and automatically segments 104 organ classes from which we pick lung lobes, trachea, pulmonary artery, and left atrium of the heart for lung segmentation. All created organ classes are converted into a single 3D Slicer segmentation with the correct naming.

The new updated version of the extension (2.53) should be available tomorrow via the extension manager.

For installation details (Windows is a bit tricky) see this wiki page:

2 Likes

Looks really cool. The TotalSegmentator repo says you need an Nvidia GPU to use it. Is that true for the Slicer extension also?

Hi Dave, 3D Slicer and this extension will run on all major operating systems with a wide range of GPU.
See the hardware requirements here.

Great!

Yeah, I use Slicer on my MacBook Pro with an AMD gpu all the time. I tried out the Lung CT Segmenter, and it produced a very nice result on the one data set I used. I just wanted to be sure about TotalSegmentator, before I try it.

1 Like