Can routine chest CT with artificial intelligence tool be used for opportunistically osteoporosis screening? A preliminary study

Introduction Patients with pulmonary diseases face an elevated risk of osteoporosis. Conventional diagnosis relies on quantitative computed tomography (QCT), which delivers a high radiation dose and requires a dedicated scan. Whether conventional chest CT can be used for osteoporosis screening remains uncertain. Methods A deep learning system was used to measure bone mineral density (BMD) in routine chest CT scans for osteoporosis screening, with QCT as the reference standard. A total of 98 patients who underwent chest and abdominal CT due to medical conditions between August 2024 and March 2025 were analyzed. The BMD of the thoracic vertebrae (T1-T12) was measured, three consecutive vertebral bodies were grouped into averages, and the results were compared with the BMD of the lumbar vertebrae (L1-L3) measured by standard QCT. The differences and agreement between the two BMD measurement methods were compared, and the diagnostic performance for osteoporosis and low bone mass was assessed by receiver operating characteristic curve analysis. Results The cohort included 98 patients (48 female, 50 male; mean age 49.48 ± 14.64 years). AI-measured thoracic vertebrae BMD correlated highly with QCT BMD (r: 0.72-0.92), with the highest correlation observed for T10-T12 (r = 0.98, P < 0.001). All AI-BMD measurements demonstrated high diagnostic performance for osteoporosis and low bone mass. In the osteoporosis analysis,the T10-T12 BMD achieved the highest AUC of 0.998 (95% CI: 0.960-1.000), with an optimal diagnostic threshold of BMD ≤90.86 mg/cm³ (sensitivity 100.0%, specificity 97.6%). In the low-bone-mass analysis, the AUC was 0.990 (95% CI: 0.945-1.000). The volumetric CT dose index for QCT scans was significantly higher than that for chest CT scans (14.63 ± 2.30 vs. 4.13 ± 1.01 mGy, P < 0.001), and the post-processing time for QCT was significantly longer than that for the AI-based measurement (55.76 ± 9.05 vs. 7.50 ± 0.78 s, P < 0.001). Conclusion AI-based BMD measurements from chest CT correlate highly with QCT measurements. AI-derived thoracic BMD from routine chest CT may serve as a potential opportunistic screening-support tool for osteoporosis and low bone mass, but further validation in larger, multicenter, prospective cohorts is required before clinical adoption.