David Dreizin, Alexis R Boscak, Michael J Anstadt, Nikki Tirada, William C Chiu, Felipe Munera, Uttam K Bodanapally, Michael Hornick, Deborah M Stein

Purpose To determine the diagnostic performance of multidetector computed tomography (CT) with trajectography for penetrating colorectal injuries. Materials and Methods This institutional review board-approved and HIPAA-compliant study was a 6-year blinded retrospective review by two independent readers of 182 consecutive patients who preoperatively underwent 40- or 64-row multidetector CT for penetrating torso trauma below the diaphragm and had surgically confirmed findings. Colorectal perforation was present in 42 patients. Trajectory analysis with postprocessing software was used for all studies. Additional signs evaluated were rectal contrast agent leak, collections of extruded fecal material, mural defect, wall thickening, abnormal enhancement, free fluid or stranding, and free air. The quality of the colorectal contrast agent administration was recorded. Sensitivity, specificity, predictive values, areas under the receiver operating characteristic curves (AUCs), and Cohen κ were determined. Results In patients with rectal contrast agent administration (n = 151), AUCs were 0.90-0.91, which indicated excellent accuracy. Trajectory was sensitive (88%-91%). For single wounds (n = 104), sensitivity of trajectory was 96% for both readers, but was only 80% for multiple wounds (n = 47). Contrast agent leak was highly specific (96%-98%), but insensitive (42%-46%). Improved diagnostic performance was observed in patients with poor colonic distension or opacification. Accuracy remained high (AUC, 0.86-0.99) in the group without rectal contrast agent administration (n = 31). Conclusion Trajectory had excellent sensitivity, while rectal contrast agent leak was specific but insensitive. Sensitivity of trajectory was lower for multiple wounds. Accuracy remained high in patients without rectal contrast agent administration. (©) RSNA, 2016.