Diagnostic performance of dual-energy computed tomography for detection of acute spinal fractures

To assess the diagnostic performance of dual-energy computed tomography (DECT) with the color-coded virtual non-calcium (VNC) technique for detecting acute fractures in patients after acute spine trauma, especially in an emergency clinical setting.

Our retrospective study included 31 patients presented to emergency department with suspected spine trauma. All patients underwent both DECT (80 kVp and 140 kVp) and MRI. Post-processing was performed using color-coded VNC technique. Two independent radiologists visually assessed color-coded VNC images in a random order, and one of the two readers re-assessed the images in 4 weeks after the initial assessment. They were allowed to read only color-coded VNC images and asked to determine the presence of acute fracture. To determine the standard reference point, the other two experienced radiologists made consensus readings on both grayscale CT and MRI. Sensitivity, specificity, PPV, NPV, and accuracy analyses were determined. Both intra- and inter-observer agreements were also calculated.

A total of 217 vertebral bodies (65 thoracic and 152 lumbar vertebrae) were included in our study. Sensitivity was 83.3% and 76.7% for first and second readers, respectively. Specificity of 99.5% and 98.9%, PPV of 96.1% and 96.3%, NPV of 97.3% and 96.3%, and accuracy of 97.2% and 95.8%, respectively, were noted. Both intra-observer and inter-observer agreements indicated excellent agreement (κ = 0.86 and κ = 0.84, respectively).

In spite of the relatively low sensitivity, DECT-based detection of acute spinal fractures showed good specificity, positive predictive value, negative predictive value, accuracy, and inter-/intra-observer agreements.