Background: The usefulness of in-phase/opposed-phase imaging and diffusion weighted imaging (DWI) in differentiating benign and neoplastic vertebral fractures has been described. In this study, we aimed to evaluate the influence of the severity of vertebral damage on the diagnostic performance of these two technologies.
Methods: Totally 59 patients with 68 acute benign vertebral fractures and 43 patients with 79 vertebral metastases were included in this study. The MR protocol included DWIs and sagittal in-phase/opposed-phase gradient recalled sequence. The severity of vertebral damage was expressed by lesion ratio (LR, the ratio of lesion area to vertebral area on the slices of largest abnormal signal area in the T1-weighted sequence). Quantitative (signal intensity ratio (SIR) defined as signal intensity (SI) on opposed-phase gradient recalled echo (GRE) images divided by SI on in-phase; apparent diffusion coefficient (ADC) value derived from DWI analysis was performed, the relationships between LR and the measurements of these two technologies were analyzed using linear regression. The covariate-specific receiver operating characteristic (ROC) curves were also fitted to evaluate the influence of LR on the diagnostic performance of ADC and SIR.
Results: The difference in both SIR and ADC for vertebral metastasis and acute benign vertebral fractures was significant (P < 0.001). A positive correlation between the LR and the SIR was found in benign fractures (P < 0.05). The severity of vertebral damage had a significant influence on the AUC (area under ROC curve) for SIR (P < 0.05) but ADC (P > 0.05). More severe cases were associated with increased AUC for SIR.
Conclusions: LR is capable of affecting the diagnostic performances of chemical shift imaging. Thus, when applying these tests to make diagnoses on vertebral fractures, the severity of the vertebral damage should be taken into account. The covariate-specific ROC model is recommended because it substantially improves the ability to avoid bias when evaluating tests.