PURPOSE: To compare the accuracy of MR imaging using an endorectal-pelvic and a pelvic phased-array coil forpreoperative local staging of rectal carcinoma. MATERIALS AND METHODS: To determine preoperative staging, 38patients with rectal carcinoma underwent MR imaging. All patients were examined with both an endorectal-pelvic anda pelvic phased-array coil. All underwent surgery and staging was pathologically confirmed. Two radiologistsblinded to pathologic stage analyzed perirectal invasion and perirectal node metastasis, and scored according to afour-point scale. Radiologic and pathologic findings were correlated. Receiver operating characteristic (ROC)analysis of Wilcoxon statistic (W values) was used to compare diagnostic accuracy between the two different MRmethods. Interobserver variation was measured using kappa statistics. RESULTS: For perirectal invasion, T1WIendorectal-pelvic phased-array coil images (reader 1: 0.854, reader 2: 0.818) showed higher W values than pelvicphased-array coil images (reader 1: 0.755, reader 2: 0.811). On T2WI, W values were higher according to pelvicphased-array coil images (reader 1: 0.828, reader 2: 0.861) than according to endorectal-pelvic phased-array coilimages (reader 1: 0.813, reader 2: 0.786). For perirectal node metastasis, pelvic phased-array coil images (reader1: 0.745, reader 2: 0.792) showed higher W values than endorectal-pelvic phased-array coil images (reader 1:0.722, reader 2: 0.775), according to both reader 1 and 2. The defference kappa values between the two readers wasless than 0.4 ; agreement between them was poor. CONCLUSION: The use of an endorectal-pelvic phased-array coildid not significantly improve the accuracy of assessment of perirectal invasion and perirectal node metastasis,and in MR imaging of rectal corcinoma, the routine use of an endorectal coil is not advocated.