Step 2 Using Lung CT Segmenter with AI tools
This page describes how the AI tool packages lungmask, MONAILabel or TotalSegmentator can be used within Lung CT Segmenter.
You will need a relatively strong computer with a medium to high end Nvidia GPU (RTX 1060 +, 6+ GB graphics memory) and CUDA enabled. Pytorch must be installed in 3D Slicer. Please refer to the SlicerPytorch extension.
General remark: If nothing else apart from "use experimental AI" is selected, the segmentation will begin immediately after pressing the "Start" button.
Using "lungmask" is relatively straightforward:
Make sure you have installed Pytorch with 3D Slicer with SlicerPytorch. Then load your lung CT in Slicer and run the Lung CT Segmenter. Select "use AI" and choose one of the "lungmask" options from the drop down box:
lungmask R231: right and left lung lungmask LTRCLobes: lobe segmentation, model trained with LTRC dataset lungmask LTRCLobes_R231: lobe segmentation, fusion lungmask R231CovidWeb: right and lung segmentation trained with additional COVID data
Details see here. The segmentations will take about 30 s to run.
TotalSegmentator is a nnU-Net-based AI tool for the segmentation of 104 classes in CT images.
Using "TotalSegmentator" within 3D Slicer is realized via the TotalSegmentator 3D Slicer extension. Please make sure the extension is installed and keep it updated.
Load your CT data in Slicer, select the Lung CT Segmenter extension, "use AI" and one of the "TotalSegmentator" options. Press "Apply". Be patient, the process may take up to 5 minutes depending on the option selected.
You can also run TotalSegmentator independently of 3D Slicer from the command line.