All datasets
No longer availableMRI

SKI10: Segmentation of Knee Images 2010 Challenge

SKI10 was a knee MRI segmentation challenge dataset containing 100 training volumes and 50 test volumes with labels for femur, femoral cartilage, tibia and tibial cartilage.

Licence and access

No licence name was ever published. The archived rules page states that the data "must not be given nor redistributed under any circumstances" and that "commercial use of segmentation algorithms that use the provided data (test or training cases) as training material is not allowed". Source: web.archive.org capture of ski10.org/rules.php, 2013-05-01.

access endedcommercial use restrictedredistribution restricted
Visit official sourcelast verified 2026-09-11This link no longer leads to the data, only to the page where the dataset was described.
Access status: Closed in 2019 after the data provider, Biomet, withdrew permission to share the dataset. The challenge page states that the SKI10 dataset can no longer be shared.
Coverage
kneesegmentationin vivo
subjects
not verified
images
150
formats
MetaImage, MHD, RAW
origin
US · 2010
Citation
Heimann T, Morrison BJ, Styner MA, Niethammer M, Warfield SK. Segmentation of Knee Images: A Grand Challenge. MICCAI 2010 Workshop on Medical Image Analysis for the Clinic. Related papers (Zenodo): 10.5281/zenodo.4781231
Notes

The MICCAI paper states that 250 knee MRI images were collected from the surgical planning program of Biomet, Inc., acquired at over 80 centres in the USA; 150 of those (100 training, 50 test) were distributed for the challenge, which is the figure recorded here. Subjects were patients undergoing planning for partial or complete knee replacement, and identifying information was removed from the DICOM headers. The Zenodo record 10.5281/zenodo.4781231 is titled "SKI10 papers" and contains 16 PDFs, not the imaging data; the image data never received a DOI, so its CC BY 4.0 licence applies to the papers only. The former ski10.org domain has been lost and re-registered: it now redirects to an unrelated site and 2025 archive captures show gambling pages, so it must not be cited as a source.

This record is a pointer. Licence and access terms are recorded as they stood on the verification date above; what binds you is the dataset's own licence text. Confirm at the source before use.

The record is kept to document that the dataset existed and where it was described.

Orthopedic AI Research Updates

Monthly research digest, product updates, and clinical AI insights.

Unsubscribe anytime.

SKI10: Segmentation of Knee Images 2010 Challenge