Accuracy of Low-Dose Chest CT-Based Artificial Intelligence Models in Osteoporosis Detection: A Systematic Review and
Huang Ya'nan1, Zhou Jianfeng2, Tang Wei1
1Department of Radiology, Shaoxing People's Hospital (Shaoxing Hospital of Zhejiang University), No. 568 Zhongxing North Road, Shaoxing, 312000, Zhejiang, China.
Calcified Tissue International
|May 2, 2025
Summary
Artificial intelligence (AI) using low-dose chest CT scans shows promise for osteoporosis screening. AI models accurately identify vertebrae and diagnose osteoporosis with high sensitivity and specificity, aiding early detection.
Area of Science:
- Radiology
- Artificial Intelligence
- Osteoporosis Research
Background:
- Osteoporosis poses a significant public health challenge, often diagnosed late.
- Low-dose chest computed tomography (CT) scans are increasingly common, offering a potential opportunity for incidental osteoporosis screening.
- Developing accurate and efficient screening tools is crucial for early intervention.
Purpose of the Study:
- To systematically review and evaluate the diagnostic accuracy of artificial intelligence (AI) algorithms applied to low-dose chest CT (LDCT) images for osteoporosis screening.
- To assess the performance of AI in vertebrae segmentation and the subsequent diagnosis of osteoporosis and osteopenia.
Main Methods:
- A systematic literature search was conducted across major databases (PubMed, Scopus, Web of Science, Cochrane Library) up to December 13, 2024.
- A meta-analysis adhering to PRISMA-DTA guidelines was performed, using modified QUADAS-2 to assess study quality.
- AI model performance was evaluated using pooled sensitivity, specificity, area under the curve (AUC), and Dice Similarity Coefficient (DSC) for vertebrae segmentation.
Main Results:
- Eight studies were included in the meta-analysis.
- The pooled Dice Similarity Coefficient (DSC) for automatic vertebrae segmentation was 0.92 (95% CI 0.88-0.97).
- For detecting abnormal (osteoporosis + osteopenia) and osteoporosis cases, pooled sensitivities were 0.90 (95% CI 0.88-0.91) and 0.86 (95% CI 0.82-0.89), respectively. Pooled specificities were 0.90 (95% CI 0.88-0.91) and 0.93 (95% CI 0.92-0.94), with summary ROC curves of 0.9653 and 0.9676.
Conclusions:
- Low-dose chest CT-based AI models demonstrate significant potential for identifying patients who may have osteoporosis or osteopenia, warranting further evaluation.
- Factors such as dataset source, annotation methods, radiomics inclusion, and spine segmentation scope influenced AI model heterogeneity.
- Prospective, multi-center, multi-dataset studies are necessary to fully establish the complementary role of AI in osteoporosis diagnosis using LDCT.
