Diagnostic value of an S1-normalized CT hounsfield unit ratio for identifying acute fractures in complex...

In patients with osteoporotic vertebral fractures, multivertebral deformities, previous vertebral fractures, adjacent vertebral augmentation, or continuous multivertebral acute fractures may make the conventional adjacent-vertebra reference unreliable. This study aimed to evaluate the diagnostic value of an S1-normalized CT Hounsfield unit ratio (S1-ratio), which does not rely on adjacent vertebrae, for identifying MRI-confirmed acute fractured vertebrae in complex thoracolumbar osteoporotic deformities.

This single-center retrospective diagnostic accuracy study included female patients who presented between January 2024 and June 2025 with low-energy T11-L5 vertebral fractures or suspected osteoporotic vertebral deformities and had complete imaging data. Target vertebrae were identified by three radiologists based on CT morphological abnormalities while blinded to MRI bone marrow edema findings; disagreements were resolved by consensus. Corresponding bone marrow edema on MRI STIR or fat-suppressed T2-weighted sequences was used as the reference standard for acute fracture. The target vertebrae and S1 vertebral body were measured on axial CT images using cancellous bone regions of interest at upper, middle, and lower levels. The mean target vertebral HU (VB-HU), S1-HU, and S1-ratio (VB-HU/S1-HU) were calculated. Receiver operating characteristic (ROC) analysis was used to evaluate diagnostic performance. Generalized estimating equation (GEE) logistic regression was used to account for within-patient correlation caused by multiple target vertebrae per patient, and vertebral region was further included to assess the effect of uneven segmental distribution.

A total of 91 female patients and 217 target vertebrae were included, comprising 100 acute fractured vertebrae and 117 chronic deformity vertebrae. VB-HU (170.51 ± 48.43 vs. 75.39 ± 32.83, P < 0.001) and S1-ratio (1.47 ± 0.53 vs. 0.66 ± 0.23, P < 0.001) were significantly higher in acute fractured vertebrae than in chronic deformity vertebrae. The AUC of S1-ratio for identifying acute fractured vertebrae was 0.968 (95% CI 0.949-0.987, P < 0.001). The optimal cutoff was 0.948, with a sensitivity of 92.0%, specificity of 89.7%, accuracy of 90.8%, positive predictive value of 88.5%, and negative predictive value of 92.9%. GEE analysis showed that each 0.1 increase in S1-ratio was associated with a 2.63-fold increase in the odds of acute fracture (95% CI 2.07-3.36, P < 0.001). After further adjustment for vertebral region, the association remained significant (OR = 2.49, 95% CI 1.93-3.22, P < 0.001). S1-HU was negatively correlated with age (rho = -0.467, P < 0.001) and positively correlated with multiple DXA T-scores. Intraobserver and interobserver reliability for VB-HU, S1-HU, and S1-ratio was good to excellent (ICC = 0.878-0.967).

S1-ratio is a simple, calculable, and non-adjacent-vertebra-dependent CT quantitative metric that may assist in identifying MRI-confirmed acute fractured vertebrae in complex osteoporotic vertebral deformities. Its diagnostic performance remained stable after accounting for within-patient correlation and uneven vertebral-region distribution, particularly in clinical scenarios where adjacent vertebrae are unsuitable references because of chronic deformity, percutaneous kyphoplasty/percutaneous vertebroplasty (PKP/PVP), instrumentation, or continuous multivertebral fractures.