Noise-Optimized Virtual Monoenergetic Dual-Energy CT Improves Diagnostic Accuracy for the Detection of Active Arterial Bleeding of the Abdomen

Purpose: To evaluate diagnostic accuracy of a noise-optimized virtual monoenergetic imaging (VMI+) reconstruction technique for detection of active arterial abdominal bleeding on dual-energy (DE) CT angiography compared with standard image reconstruction. Materials and Methods: DE CT angiography data sets of 71 patients (46 men; age 63.6 y +/- 13.3) with suspected arterial bleeding of the abdomen or pelvis were reconstructed with standard linearly blended (F_0.5), VMI+, and traditional virtual monoenergetic imaging (VMI) algorithms in 10-keV increments from 40 to 100 keV. Attenuation measurements were performed in the descending aorta, area of hemorrhage, and feeding artery to calculate contrast-to-noise ratios (CNRs) in patients with active arterial bleeding. Based on quantitative image quality results, the best series for each reconstruction technique were chosen to analyze the diagnostic performance of 3 blinded radiologists. Results: DE CT angiography showed acute arterial bleeding in 36 patients. Mean CNR was superior in 40-keV VMI+ compared with VMI series (all P < .001), which showed highest CNRs in 70-keV VMI and F_0.5 (21.6 +/- 7.9, 12.9 +/- 4.7, and 10.4 +/- 3.6) images. Area under the curve analysis for detection of arterial bleeding showed significantly superior (P < .001) results for 40-keV VMI+ (0.963) compared with 70-keV VMI (0.775) and F_0.5 (0.817) series. Conclusions: Diagnostic accuracy in patients with active arterial bleeding of the abdomen can be significantly improved using reconstructions at 40 keV compared with standard linearly blended and traditional VMI series in DE CT angiography.