Diagnostic Accuracy of 3.0-Tesla Rectal Magnetic Resonance Imaging in Preoperative Local Staging of Primary Rectal Cancer

Abstract

To evaluate the diagnostic accuracy of 3.0-T rectal magnetic resonance imaging (MRI) in the preoperative local staging of primary rectal cancer. Forty-two patients with surgically and pathologically proven primary rectal cancer who underwent preoperative gadobenate dimeglumine-enhanced 3.0-T rectal MRI, were enrolled in this retrospective study. Two radiologists, who were blinded to the pathology results, independently reviewed the MR images and recorded their confidence level for determination of perirectal extension, and regional lymph node (LN) involvement using a 5-point scale. The diagnostic accuracy of each reviewer for local staging was calculated by receiver operating characteristic (ROC) curve analysis. Interobserver agreement was also calculated using linear weighted kappa statistics. The diagnostic accuracy (area under the ROC curve, Az) for determining perirectal extension was for reviewer 1, 0.860 (95% confidence interval, 0.72-0.95) and for reviewer 2, 0.853 (0.71-0.94), respectively. The Az for determination of regional LN involvement was for reviewer 1, 0.902 (0.77-0.97) and for reviewer 2, 0.843 (0.70-0.94), respectively. Interobserver agreement included, respectively, good, and moderate agreement for perirectal extension, and regional LN involvement (kappa = 0.662, and 0.522, respectively). 3.0-T rectal MRI can provide accurate information of perirectal extension and regional LN involvement in the preoperative local staging of primary rectal cancer.