The accuracy of 64-slice MDCT for determining the perforation site of the gastrointestinal tract: ROC analysis

Purpose: This study was performed to evaluate the diagnostic performance of 64-slice multidetector-row computed tomography (MDCT) for preoperatively detecting the perforation site in patients with gastrointestinal (GI) tract perforations. Materials and methods: A total of 49 patients with pathologically proven GI tract perforations, who had undergone MDCT were included in this study. The contrast-enhanced axial images (3 mm thick) and multiplanar reformation (MPR) images (3 mm thick) were generated for all the patients. Two radiologists independently reviewed the two sets of MDCT images (axial set: the axial images alone, and combined set: the axial and MPR images) for the detection of the perforation site. The perforation site was considered to be positive according to the direct and indirect findings. The diagnostic accuracy was assessed with a receiver operating characteristic (ROC) analysis. Weighted kappa statistics were used to evaluate the interobserver agreement. Results: The corresponding values in respect of the first and the second observers for the areas under the ROC curve were 0.984 and 0.966 for the axial set; and 0.998 and 0.973 for the combined set. The differences were not statistically significant between the two data sets for each observer (P > 0.05). Both the observers detected the perforation site in 43 (87.8%) and 40 (81.6%) patients on the axial set, and in 46 (93.9%) and 41 (83.7%) patients on the combined set, respectively. The kappa values between the two observers showed excellent agreement with the two data sets. Conclusion: 64-slice MDCT showed a consistently excellent diagnostic performance for preoperatively determining the perforation site in patients with GI tract perforations, irrespective of the availability of the MPR images.