Deep learning on routine chest radiographs predicts major osteoporotic fractures better than FRAX
A convolutional survival model trained on 53,109 adults at a Korean tertiary hospital and validated on 85,284 patients from two other institutions predicted incident major osteoporotic fracture from routine chest radiographs with a C-index of 0.843, against 0.779 for FRAX with bone density.
- Model
- A convolutional neural network survival model; not named
- Field
- Medicine
- Date
- 2026-11-16
- Human collaborators
- Yisak Kim, Chang Min Park, Hanbin Ko, Dongheon Lee, Jin Hwan Kim, Sung Hye Kong
What was found
Bone density and FRAX miss much of the fracture risk in the osteopenic range, where most fractures occur. Participants needed a chest radiograph within a year of a DXA scan. In the external cohorts (mean age 58.6, 68.9% female, 77 months of follow-up) the model reached a C-index of 0.843 (95% CI 0.837 to 0.850) and a censoring-adjusted 10-year AUROC of 0.855, against 0.779 and 0.785 for FRAX with BMD (both ); adding FRAX to the model barely changed it (C-index 0.847). In the osteopenia subgroup of 35,358 patients the 10-year AUROC was 0.779 against 0.659, and the advantage held across fracture types and BMD categories. The study was funded by the National Research Foundation of Korea and others.
Novelty check
Europe PMC, 2026-11-19: the DOI returns the NEJM AI record, published 16 November 2026. Opportunistic osteoporosis screening from chest radiographs and CT with deep learning is an established line, so the new element is a survival model for incident fracture compared head to head with FRAX on external cohorts, not the idea of reading bone health from chest films. Nothing in this registry covers osteoporosis or fracture prediction.
Caveats and known objections
Retrospective. Every participant had been referred for densitometry, so the cohort is not the unselected routine-radiography population the method is meant for, and the authors themselves say absolute-risk and case-finding performance needs prospective evaluation there. The abstract calls the two validation institutions independent but does not name them; the author list spans Seoul National University Hospital, Seoul National University Bundang Hospital and Chungnam National University, so the validation is not treated as one by unaffiliated investigators in this registry's sense. The outcome, incident fracture, is a hard endpoint, but discrimination is not the same as showing that acting on the score prevents fractures. Model weights and data access were not checked. Autonomy is ai-assisted: the model produces a risk score that clinicians would act on.
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 on routine chest radiographs predicts major osteoporotic fractures better than FRAX. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-16-chest-radiograph-fracture-risk
BibTeX
@misc{whataifound-seoulnationaluniversityhospital-2026-risk,
title = {Deep learning on routine chest radiographs predicts major osteoporotic fractures better than FRAX},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Seoul National University Hospital. Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-11-16-chest-radiograph-fracture-risk}
}