Dual-energy CT in vertebral compression fractures: performance of visual and quantitative analysis for bone marrow edema demonstration with comparison to MRI

To prospectively evaluate the performance of virtual non-calcium (VNC) dual-energy CT (DECT) images for the demonstration of trauma-related abnormal marrow attenuation in collapsed and non-collapsed vertebral compression fractures (VCF) with MRI as a reference standard. Twenty patients presenting with non-tumoral VCF were consecutively and prospectively included in this IRB-approved study, and underwent MRI and DECT of the spine. MR examination served as a reference standard. Two independent readers visually evaluated all vertebrae for abnormal marrow attenuation ("CT edema") on VNC DECT images; specificity, sensitivity, predictive values, intra and inter-observer agreements were calculated. A last reader performed a quantitative evaluation of CT numbers; cut-off values were calculated using ROC analysis. In the visual analysis, VNC DECT images had an overall sensitivity of 84 %, specificity of 97 %, and accuracy of 95 %, intra- and inter-observer agreements ranged from k = 0.74 to k = 0.90. CT numbers were significantly different between vertebrae with edema on MR and those without (p < 0.0001). Cut-off values provided sensitivity of 85 % (77 %) and specificity of 82 % (74 %) for "CT edema" on thoracic (lumbar) vertebrae. VNC DECT images allowed an accurate demonstration of trauma-related abnormal attenuation in VCF, revealing the acute nature of the fracture, on both visual and quantitative evaluation.