Objective:

To explore methods for osteoporosis detection using hip CT data analysis, integrating 2D/3D deep learning, radiomics, and clinical data to improve diagnostic accuracy and reliability.

Approach:

• Study Population: 567 patients with hip joint CT images were enrolled.
• Data Integration: Clinical data including age and gender were used to establish a clinical model for opportunistic osteoporosis screening.
• Model Development: Regions of interest on hip joint CT scans were assessed using radiomic techniques, 2D and 3D deep learning technologies.
• Nomogram Model: A Nomogram model for opportunistic osteoporosis screening was established by integrating the radiomic model with the clinical model.
• Model Comparison: The efficacy of each model was compared to identify the optimal model for screening.

Key Findings:

• The GradientBoosting machine learning algorithm demonstrated the best performance in the validation group, achieving an accuracy of 0.849 and an AUC of 0.911.
• Among 2D deep learning models, densenet201 achieved an accuracy of 0.817 and an AUC of 0.884 in the validation group.
• In 3D deep learning models, ResNet34 showed an accuracy of 0.806 and an AUC of 0.889 in the validation group.

Interpretation:

The study presents findings on the integration of radiomics and deep learning techniques for evaluating osteoporosis through hip CT imaging.

Conclusion:

The study demonstrates the effectiveness of multimodal approaches in opportunistic osteoporosis detection using hip CT imaging.