Deep learning detection and staging of femoral head osteonecrosis on hip x-rays across six Chinese hospitals
OsteoAID, a YOLOv8-based model trained on 20,722 hip x-rays from 16,399 patients at six Chinese hospitals, staged osteonecrosis of the femoral head with 96.6% accuracy on an external test set, against 75.0% for radiologists.
- Model
- OsteoAID (YOLOv8-based)
- Field
- Medicine
- Date
- 2026-10-26
- Human collaborators
- Zhanqi Wei, Xisheng Weng
What was found
Expert orthopaedists annotated each image to separate osteonecrosis from radiographically similar conditions and graded confirmed lesions on the Ficat and Arlet scale, using images from 2015 to 2023. On the internal test set the model raised staging accuracy from 87.4% for the unmodified YOLOv8 network to 96.4%, with a detection mean average precision of 93.0% at an intersection-over-union threshold of 50%. It was compared with radiologists and orthopaedists on two external test sets, and with MRI-based staging, the reference standard, on a third. The authors propose it as a decision-support tool for triage.
Novelty check
Europe PMC, 2026-10-29: TITLE_ABS:"osteonecrosis of the femoral head" AND TITLE_ABS:"deep learning" returned 28 records, most on MRI, including 2026 systematic reviews of deep learning for the condition. Deep learning for this diagnosis is established; what is distinctive here is staging on plain radiographs at this scale with clinician comparison on external sets.
Caveats and known objections
Retrospective and multicentre, run by the developers at tertiary hospitals in China, and the abstract does not say how independent the external sets are from the training institutions. Funded by the National Natural Science Foundation of China and other public Chinese sources; no vendor is named. The endpoint is diagnostic accuracy against expert annotation, with MRI as reference on one set, a surrogate rather than a treatment outcome, and the headline comparison is of the model alone against clinicians alone. Whether the model or data are available outside the group is not stated. Autonomy is ai-assisted: the model is proposed as decision support for clinicians.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Deep learning detection and staging of femoral head osteonecrosis on hip x-rays across six Chinese hospitals. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-26-osteoaid-femoral-head-osteonecrosisBibTeX
@misc{whataifound-sunyatsenuniversity-2026-osteonecrosis,
title = {Deep learning detection and staging of femoral head osteonecrosis on hip x-rays across six Chinese hospitals},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Sun Yat-sen University (with the Chinese Artificial Intelligence Project Group for Hip Joint Diseases). Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-10-26-osteoaid-femoral-head-osteonecrosis}
}