The purpose of this study was to evaluate the accuracy of 3-Tesla magnetic resonance imaging (MRI) for the preoperative staging of rectal cancer. Thirty-five patients with a primary rectal cancer who underwent preoperative 3-T MRI using a phased-array coil and had a surgical resection were enrolled in the study group. Preoperatively, three experienced radiologists independently assessed the T and N staging. A confidence level scoring system was used to determine if there was any perirectal invasion, and receiver operating characteristic (ROC) curves were generated. The interobserver agreement was estimated using κ statistics. The overall accuracy rate of T staging for rectal cancer was 92%. The diagnostic accuracy was 97% for T1, 89% for T2 and 91% for T3, respectively. The predictive accuracy for perirectal invasion by the three observers was high (Az>0.92). The interobserver agreement for T staging was moderate to substantial. The overall sensitivity, specificity, and accuracy for the detection of mesorectal nodal metastases were 80%, 98%, and 95%, respectively. In conclusion, preoperative 3-T MRI using a phase-array coil accurately indicates the depth of tumor invasion for rectal cancer with a low variability.