Detection of intestinal bleeding with multi-detector row CT in an experimental setup. How many acquisitions are...

The purpose of the study was to evaluate multi-detector computed tomography (MDCT) acquired in different acquisitions (unenhanced, and arterial and portal venous phase following intravenous contrast medium) for detection of intestinal bleeding using an experimental bowel model. The model consisted of an injector tube with a perforation placed in a 7-m-long small bowel of a pig. The bowel was filled with water/contrast medium solution of 30–40 HU and was incorporated in a phantom model. Intestinal bleeding in different locations and bleeding velocities varying from zero to 0.75 ml/min (0.05 ml/min increments) were simulated. Twenty-six datasets in simulated unenhanced, arterial and portal venous contrast phase using increasing bleeding velocities and ten negative controls were measured using 64-row MDCT. Two radiologists blinded to the experimental settings evaluated the datasets in a random order. The likelihood of intestinal bleeding was assessed using a 5-point scale with subsequent ROC analysis. The overall sensitivity for detecting bleeding was 0.44 for an arterial acquisition alone, 0.68 for a portal venous acquisition, 0.68 for the combination unenhanced/arterial, 0.72 for unenhanced/portal venous and 0.80 for arterial/portal. Bleeding velocities of above 0.25 ml/min were detected with a sensitivity of 0.59 for arterial, 0.88 for portal venous, 0.85 for unenhanced/arterial, 0.94 for unenhanced/portal venous and 0.97 for arterial/portal venous contrast phase protocols, respectively. The specificity was 1.00. MDCT provides the highest sensitivity and specificity in the detection of intestinal bleeding using arterial and portal venous acquisition in comparison to mono-phase protocols.