All datasets
AvailableCT

DeepLesion

DeepLesion is a large clinical CT dataset containing lesions marked during routine radiology interpretation. It includes RECIST-style lesion measurements and bounding boxes across multiple body regions, including bone, spine and pelvic lesions.

Licence and access

DeepLesion is publicly distributed by the NIH Clinical Center for research use. The download page does not provide a conventional open-data license that clearly authorizes commercial reuse, so commercial use is recorded as unknown rather than assumed.

open accesscommercial use unverifiedredistribution allowed
Visit official sourcelast verified 2026-09-14
Coverage
spinepelvistumorothermixed labelsin vivo
subjects
4,427
images
32,120
formats
PNG, CSV
origin
not verified
Citation
Yan K, Wang X, Lu L, Summers RM. DeepLesion: Automated Mining of Large-Scale Lesion Annotations and Universal Lesion Detection with Deep Learning. Journal of Medical Imaging. 2018;5(3):036501. DOI: 10.1117/1.JMI.5.3.036501.
Notes

The final public release contains 32,735 lesions on 32,120 axial CT slices from 10,594 studies belonging to 4,427 patients. The collection is not exclusively musculoskeletal; it also contains lung, liver, lymph-node, abdominal and other lesions. It is included because identifiable bone and soft-tissue lesion groups are represented. The released images are stored as 16-bit PNG files with annotations and metadata in CSV tables. The only usage statement NIH publishes is FAQ item 7, "The usage of the data set is unrestricted"; the official Box folder carries no separate licence or terms file. Commercial use is left as not verified because an FAQ entry is not a licence.

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.

Orthopedic AI Research Updates

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

Unsubscribe anytime.

DeepLesion, dataset record