OBJECTIVE: The aim of this study was to compare phased-array 3-T MRI and endorectal sonography in the preoperative staging of rectal cancer. MATERIALS AND METHODS: During an 8-month period, 24 patients with rectal cancer underwent both 3-T MRI performed with phased-array coils and 7.5- to 10-MHz endorectal sonography in the 3 weeks before surgical resection. Three radiologists independently reviewed the MR and endorectal sonographic images. The histopathologic findings in resected specimens were used to evaluate the sensitivities and specificities of these techniques for invasion of the muscularis propria and perirectal tissue and for lymph node involvement. Receiver operating characteristic (ROC) analysis was used to compare the diagnostic accuracies of the techniques. RESULTS: For muscularis propria invasion, the mean sensitivities of both MRI and endorectal sonography were 100%, and the mean specificities were 66.7% and 61.1%, respectively. The differences in the mean sensitivities and specificities were not statistically significant (p > 0.05 in each case). For perirectal tissue invasion, MRI and endorectal sonography had comparable sensitivities and specificities (91.1% vs 100%, 92.6% vs 81.5%; p > 0.05 in each case). They also had similar sensitivities and specificities for lymph node involvement (63.6% vs 57.6%, 92.3% vs 82.1%; p > 0.05 in each case). ROC curves for muscularis propria invasion and lymph node involvement showed no differences in diagnostic accuracy. The mean area under the ROC curve for endorectal sonography (A(Z) = 0.996) for perirectal tissue invasion, however, showed higher accuracy than that of MRI (A(Z) = 0.938, p = 0.028). CONCLUSION: The sensitivity, specificity, and accuracy of 3-T MRI were similar to those of endorectal sonography for muscularis propria invasion and lymph node involvement, but for perirectal tissue invasion, 3-T MRI was less accurate than endorectal sonography.