Dual-source dual-energy CT angiography with virtual non-enhanced images and iodine map for active gastrointestinal bleeding: Image quality, radiation dose and diagnostic performance

Highlights: • GIB is a common gastrointestinal emergency with a high mortality rate. • Detection and localization of GIB source are important for imaging modality. • DSDECTA using a dual-phase scan protocol is clinically feasible. • DSDECTA with VNE and iodine map images can diagnose the active GIB source accurately. • DSDECTA can reduce radiation dose compared with conventional CT examination in GIB. - Abstract: Objectives: To evaluate the clinical feasibility of dual-source dual-energy CT angiography (DSDECTA) with virtual non-enhanced images and iodine map for active gastrointestinal bleeding (GIB). Methods: From June 2010 to December 2012, 112 consecutive patients with clinical signs of active GIB underwent DSDECTA with true non-enhanced (TNE), arterial phase with single-source mode, and portal-venous phase with dual-energy mode (100 kVp/230 mAs and Sn 140 kVp/178 mAs). Virtual non-enhanced CT (VNE) image sets and iodine map were reformatted from 'Liver VNC' software. The mean CT number, noise, signal to noise ratio (SNR), image quality and radiation dose were compared between TNE and VNE image sets. Two radiologists, blinded to clinical data, interpreted images from DSDECTA with TNE (protocol 1), and DSDECTA with VNE and iodine map (protocol 2) respectively, with discordant interpretation resolved by consensus. The standards of reference included digital subtraction angiography, endoscopy, surgery, or final pathology reports. Receiver–operating characteristic (ROC) analysis was undertaken and the area under the curve (AUC) calculated for CT protocols 1 and 2, respectively. Results: There was no significant difference in mean CT numbers of all organs (including liver, pancreas, spleen, kidney, abdominal aorta, and psoas muscle) (P > 0.05). Lower noise and higher SNR were found on VNE images than TNE images (P < 0.05). Image quality of VNE was lower than that of TNE without significant difference (P > 0.05). The active GIB source was identified in 84 patients, 83 (83/84, 98.8%) of which were confirmed by one or more reference standard. The AUC was 0.935 ± 0.027 and 0.947 ± 0.026 for protocols 1 and 2, respectively. There was no significant difference between protocols 1 and 2 for diagnostic performance (Z = 1.672, P > 0.05). The radiation dose reduction achieved by omitting the TNE acquisition was (30.11 ± 6.32)%. Conclusion: DSDECTA with arterial phase with single-source mode, portal-venous phase with dual-energy mode and post-processing VNE image sets and iodine map could act as an accurate screening method for detection and localization of active GIB with lower radiation dose