Combination of artificial intelligence and chest computed tomography to assess bone mineral density.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou, China.
- Department of Nephrology, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou, China.
- Statistical Office, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou, China.
- Department of Nephrology, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou, China. [email protected].
- Department of Radiology, Zhangzhou Affiliated Hospital of Fujian Medical University, Zhangzhou, China. [email protected].
Abstract
To evaluate the diagnostic accuracy of artificial intelligence-assisted opportunistic chest CT for osteoporosis/osteopenia screening in a Chinese population. This retrospective study included 1306 adults (≥ 55 years) undergoing concurrent chest CT and DXA during physical examinations (Apr 2015 to May 2024). Exclusion criteria comprised vertebral fractures, spinal surgery, or contrast-enhanced CT. DXA T-scores (lumbar spine) defined osteoporosis, osteopenia, and normal BMD. The AI system automatically quantified volumetric BMD (AI-BMD) at vertebrae T10-L1. Diagnostic performance was assessed using ROC curves. Mean age was 63.96 years (36.98% male). Osteoporosis prevalence was 36.45% (n = 476/1306), significantly higher in women (P < 0.01). AI-BMD values decreased significantly from T10 to L1 (P < 0.05). For osteoporosis detection, AI-BMD demonstrated excellent diagnostic accuracy: AUC was 0.84 (95% CI: 0.81-0.87) at T10, 0.83 (95% CI: 0.80-0.86) at T11, 0.81 (95% CI: 0.79-0.84) at T12, and 0.83 (95% CI: 0.80-0.87) at L1. Performance for osteopenia diagnosis was moderate, with AUCs ranging from 0.72 to 0.75. The combination of opportunistic chest CT and AI shows promise for accurate osteoporosis screening (AUCs 0.81-0.84), albeit with moderate performance in detecting osteopenia (AUCs 0.72-0.75). The retrospective, single-center design of this study suggests that future multi-center validation is warranted to confirm the generalizability of these findings.