Skip to content
Open Imaging Index
CT AbdomenKidney

KNIGHT

KNIGHT Challenge - Kidney clinical Notes and Imaging to Guide and Help personalize Treatment and biomarkers discovery

Preoperative contrast CT and clinical records of kidney tumor patients, labelled with the postoperative American Urological Association risk group to predict before surgery who is a candidate for adjuvant therapy. Reuses the KiTS21 training cases.

Go to dataset Edit Report

Overview

KNIGHT was a challenge held with the 2022 IEEE International Symposium on Biomedical Imaging (ISBI). Teams predicted, from preoperative CT and clinical information, the risk group that the American Urological Association (AUA) guidelines assign to a renal mass after surgery. It adds no new public images: the released cases are the 300 training patients of the KiTS21 kidney tumor segmentation challenge, with a new risk label and a curated clinical record per patient. The images and segmentations of these cases are also part of KiTS23.

Composition

The development set holds 300 patients. A JSON file in the KNIGHT repository gives each patient's clinical data, including age at nephrectomy, sex, body mass index, comorbidities, smoking and alcohol history, preoperative kidney function (eGFR), radiographic tumor size, surgical details, pathology and survival. The organizers evaluated submissions on a further 103 patients selected by the same criteria; these were not released in the repository.

Acquisition

Patients had partial or radical nephrectomy for a renal tumor between 2011 and 2020, at the University of Minnesota Fairview medical center or the Cleveland Clinic. Each case is one late arterial phase abdominal CT. Imaging was fetched as NIfTI with a download script.

Annotations

The target attribute is the AUA risk group with five levels: benign, low, intermediate, high and very high risk. The main task groups high and very high risk, the candidates for adjuvant therapy, against the three lower groups; a second task predicts all five levels. Only twelve clinical attributes, such as age, sex, BMI, comorbidities, smoking, preoperative eGFR, tumor size and voxel spacing, may be used as model inputs. The KiTS21 kidney and tumor segmentations of the same cases can be taken from the KiTS21 repository.

Known limitations

  • The download script points to a storage bucket that no longer exists; the images are now obtained through the KiTS repositories.
  • The KNIGHT repository's MIT license covers the clinical file; the imaging is distributed by KiTS under CC BY-NC-SA 4.0.
  • Counting high and very high risk in the clinical file gives 85 positive cases, while the outcome paper reports 87.
  • No test labels or test images are public.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 403 subjects.

Split

  • Training 300 74%
    Age mean 58 ± 15
  • Test 103 26%
    Age mean 63 ± 12

From neheller/KNIGHT GitHub repository README (data access, prediction target, inference-time attributes); Methods, Results in Barros et al. 2025, PLOS ONE, preoperative kidney tumor risk estimation (KNIGHT challenge outcomes)

Sex by split

Reported cross table. A dot marks a cell the source does not give.

Training
Female120
Male179

gender in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case

Condition by split

Reported cross table. A dot marks a cell the source does not give.

Training
Kidney cancer275

malignant in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case

Age by split

Reported cross table. A dot marks a cell the source does not give.

Training
0-2911
30-3915
40-4938
50-5978
60-6997
70-7948
80+13

age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case

Contrast / sequence by split

Reported cross table. A dot marks a cell the source does not give.

Training
Arterial phase CT300

Methods in Barros et al. 2025, PLOS ONE, preoperative kidney tumor risk estimation (KNIGHT challenge outcomes)

Other reported numbers 8
Subjects · Training, Female, 40-49 13gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Male, 40-49 25gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Female, 50-59 30gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Male, 50-59 48gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Female, 60-69 40gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Male, 60-69 57gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Female, 70-79 18gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case
Subjects · Training, Male, 70-79 29gender, age_at_nephrectomy in knight/data/knight.json in neheller/KNIGHT (MIT), clinical data of the 300 training cases, counted per case

License and access

Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer

Access
Open download

Download without an account

Access page

Different parts of the data carry different licenses. The summary on the right shows the most restrictive answer per rule.

For Clinical data and risk group labels (knight/data/knight.json) and the download script in the KNIGHT GitHub repository.

MIT License

Use, copy, modify and redistribute the material for any purpose, including commercially, as long as the copyright notice and the permission notice are kept in all copies.

Original license text Version read: MIT License (SPDX MIT) Checked 2026-10-11

What you can do

  • Yes
  • Yes
  • Yes
  • Yes
  • Yes

What you can share

  • Yes
  • Yes
  • Yes

What you must do

  • Yes
  • Share alike No
  • No
  • No
  • Manuscript review No
  • Release code No
  • Return results No
  • Delete after use No

Limits

  • No
  • Location limits No
For CT imaging of the same KiTS case ids, as distributed through the Hugging Face repository neheller/KiTS-Challenge-Imaging that the KiTS repositories download from.

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International

Non-commercial use only, with credit. Anything you share that is built on the data must use the same license.

The KNIGHT repository states no license for the imaging. Its download script points to a storage bucket that no longer exists.

Original license text Version read: 4.0 Checked 2026-10-07

What you can do

  • No
  • Not stated
  • Conditional
  • Yes
  • Yes

What you can share

  • Conditional
  • Conditional
  • Conditional

What you must do

  • Yes
  • Yes
  • No
  • No
  • Manuscript review No
  • Release code No
  • Return results No
  • Delete after use No

Limits

  • No
  • Location limits No

Citation

Barros V, Abdallah N, Ozery-Flato M, et al. Preoperative kidney tumor risk estimation with AI: From logistic regression to transformer. PLOS ONE (2025). doi:10.1371/journal.pone.0323240

Sources

Every number on this page comes from one of these documents. Each chart names the table or page it is taken from. The raw numbers are in stats.csv.