Artificial intelligence predicted osteoporosis from chest X-rays
In simulations, prescreening could reduce bone-density scans by about 50% while retaining detection of 98–99% of osteoporosis cases.
122,535 participants, cross-sectional study, independent replication
In a computational study, researchers developed an artificial intelligence model using paired chest X-rays and bone-density scans from 26,757 patients at a tertiary referral hospital. They used 46,478 paired examinations collected between 2016 and 2022. The model predicted osteoporosis, a condition that weakens bones, and reduced bone mineral density—the amount of minerals in bone.
They tested the model in three independent cohorts: 6,391 people in a hospital cohort and 24,132 and 122,535 in two health-check cohorts. A standard measure of how well it distinguished osteoporosis cases from non-cases gave scores of 0.88, 0.93 and 0.92, respectively. Simulations suggested prescreening could also retain detection of 80–90% of reduced bone-density cases while reducing scan use.
Why it matters
Osteoporosis is an age-related condition that weakens bones. The study addresses whether chest X-rays may help identify who needs dedicated bone-density testing.
Caveats
The estimated reduction in scan use came from simulations, not observed changes in clinical care. Predictive performance alone does not establish whether prescreening improves health outcomes.
The paper
Artificial intelligence for opportunistic osteoporosis screening on chest radiographs
Seoul National University
NPJ Digital Medicine, 9 Oct 2026, CC BY-NC-ND
