Coronal CT Attenuation Measurement for Osteoporosis Screening at the Proximal Femur: A Comparative Study with the Axial Approach.
Kaifeng Ye, Junbo Qi, Jixing Fan, Yutian Luo, Yanlei Dong, Yuanyu Hu, Zhuo Chen, Yun Tian
Abstract
Open AccessObjectives: There is currently a lack of research on coronal Computed Tomography (CT) attenuation measurements. This study aimed to evaluate a coronal CT attenuation measurement method for osteoporosis screening at the proximal femur and compare its performance with the conventional axial approach. Methods: A retrospective analysis was conducted on 708 patients who underwent both proximal femur CT and dual-energy X-ray absorptiometry (DXA) within a 6-month period between January 2013 and December 2023. The axial and coronal CT attenuation values for these patients were measured. The correlation, agreement, and efficacy in screening osteoporosis of the two methods were further evaluated and compared. Results: Both measurement methods demonstrated excellent inter-observer reliability. Axial measurements yielded slightly higher HU values than coronal measurements (mean difference: 3.93 Hounsfield unit (HU), p < 0.001), and the Bland-Altman analysis showed a Coefficient of Repeatability (CR) = 9.09 HU (95% CI: 8.63 to 9.61) between the two methods. A strong correlation was observed between the two measurements (Pearson's coefficient (r) of 0.919 (95% CI: 0.907-0.930, p < 0.001)). Coronal measurements showed comparable correlation with femoral T-scores to axial measurements (0.766 (95% CI: 0.730-0.808) vs. 0.736 (95% CI: 0.697-0.771), p = 0.195). Coronal and axial measurements exhibit good predictive performance for osteoporosis diagnosis, with no statistically significant difference in AUC values between the two methods (0.850 vs. 0.864, p = 0.097). Conclusions: The coronal CT attenuation measurement method provides a reliable and complementary diagnostic tool for opportunistic osteoporosis screening, demonstrating strong correlation with axial measurements and comparable diagnostic accuracy.