A novel diagnostic approach for acute vertebral fractures based on relative Hounsfield unit values

To evaluate the diagnostic performance of Hounsfield unit (HU) values and intrapatient relative HU values in detecting acute vertebral fractures (AVFs). Data including age, sex, fracture levels, and HU values of vertebrae from T10 to L5, were analyzed. HU values were measured by placing regions of interest within the vertebral body on axial computed tomography (CT) images at the pedicle level. The HU values of the fractured vertebrae (FV) were compared with those of the nonfractured vertebrae (NFV). The FV-relative value (FV-RV) was calculated as the HU value of FV divided by the mean HU value of adjacent cranial and caudal NFV. The NFV-relative value (NFV-RV) was calculated using the same approach as the FV-RV. Comparisons were performed between FV-RV and NFV-RV. Statistical analyses included the Mann–Whitney U test, Spearman’s rank correlation, and receiver operating characteristic (ROC) curve analysis to determine optimal cutoff values for detecting AVFs. A total of 109 patients with AVFs were included in this study. Among the 855 vertebrae evaluated, 146 were fractured. The mean HU values of FV were significantly higher than those of NFV (168.3 ± 76.1 vs 91.1 ± 52.7, P < 0.001). Furthermore, FV-RV were significantly higher than NFV-RV (2.09 ± 1.51 vs 1.01 ± 0.22, P < 0.001). The ROC curve analysis revealed an optimal HU cutoff of 108.6, whereas the optimal FV-RV cutoff was 1.21, with FV-RV demonstrating better diagnostic performance compared with HU values alone. An FV-RV > 1.21 enabled AVF detection, and this CT-based intrapatient relative HU metric provides a simple adjunct for AVF diagnosis when MRI cannot be readily performed.