KiTS19 — 2019 Kidney Tumor Segmentation Challenge (re-host)

Re-host of the KiTS19 training set (210 cases with public ground-truth masks) in the same dataset/case_XXXXX/ + train.jsonl layout as KiTS23 / SLIVER07 so a single Base3DDataset subclass can load it. Imaging from the official neheller/KiTS-Challenge-Imaging; segmentations from the official neheller/kits19 GitHub repo.

Composition

Split	Cases	With mask
train	210	yes

The 90 KiTS19 test cases (case_00210..case_00299) have masks withheld by the challenge and are not included here. case_00000..case_00209 are the public training cases.

Mask labels

Value	Class
0	background
1	kidney
2	tumor

Files

dataset/case_00000/imaging.nii.gz        # contrast-enhanced abdominal CT
dataset/case_00000/segmentation.nii.gz   # 0/1/2 label volume
...
train.jsonl

License

Data is CC BY-NC-SA 4.0 (non-commercial), per the KiTS19 challenge terms. The repo code license (MIT) does not cover the data. Cite:

@article{heller2019kits19,
  title   = {The KiTS19 Challenge Data: 300 Kidney Tumor Cases with Clinical Context,
             CT Semantic Segmentations, and Surgical Outcomes},
  author  = {Heller, Nicholas and Sathianathen, Niranjan and Kalapara, Arveen and others},
  journal = {arXiv preprint arXiv:1904.00445},
  year    = {2019}
}