CT data for training this model is available on dropbox here in nrrd format.
It is originally from the TCIA CT Lymph Nodes collection in the orginal DICOM.
After downloading the server can be started like this (assumes MONAI Label is already installed and working in your python environment).
monailabel start_server --app apps/radiology --studies datasets/tcia-ln10/imagesTr --conf models segmentation
Note: this is a work-in-progress model that does not yet segment lymph nodes correctly.
This link contains three versions of the data run through TotalSegmentator and then combined with the manual segmentations from the CT Lymph Node collecion:
https://www.dropbox.com/sh/ish3ci1h9zkbbhu/AAC84tfHmQ5whW0HF-bqe5vGa?dl=0
The three subdirectories contain:
- tcia-totalseg-ln The full TotalSegmentator segmentations with the Lymph Node segment added as label 255
- tcia-totalseg-ln2 All TotalSegmentator segments are collapsed to the label value 1, and lymph nodes are label value 2
- tcia-totalseg-ln10 TotalSegmentator classes reduced to 10 label values according to the mapping below and lymph nodes are label 11
Mapping for ln10 data:
segmentMaps = [
[1, 12, 1], # organs
[13, 17, 2], # lungs
[18, 41, 3], # spine
[42, 49, 4], # cardio-pulmonary
[50, 50, 5], # brain
[51, 57, 6], # guts
[58, 87, 7], # bones
[88, 92, 8], # hips
[93, 93, 9], # face
[94, 103, 10], # muscles
[104, 104, 6], # bladder
]
All segmentations are in nii.gz format.
This work is supported by the US National Institutes of Health, National Cancer Institutes grant 5R01CA235589 Lymph Node Quantification System for Multisite Clinical Trials . Additional support provided by the NCI Imaging Data Commons.