A prospective study was designed to compare transvaginal sonography with contrast-enhanced MR imaging to determine preoperatively the depth of myometrial invasion in patients with early-stage endometrial carcinoma. In 40 patients, findings on transvaginal sonograms, unenhanced T2-weighted MR images, and contrast-enhanced T1-weighted Mr images were compared with histologic findings. The depth of myometrial invasion was classified as stage E (tumor limited to endometrium, n = 12), stage S (superficial invasion: tumor invades up to 50% of the myometrium, n = 15), or stage D (deep invasion: tumor invades more than 50% of the myometrium, n = 13). Findings on transvaginal sonograms were accurate in 27 of 40 patients (accuracy, 68%); the depth of invasion was overestimated in five patients and underestimated in eight patients. The results of unenhanced T2-weighted MR images were accurate in 27 patients (accuracy, 68%), with four overestimations and nine underestimations. The results of contrast-enhanced T1-weighted MR images were accurate in 34 patients (accuracy, 85%), with five underestimations and one overestimation. In the assessment of each stage of myometrial invasion, the sensitivity and specificity of contrast-enhanced T1-weighted imaging were higher than those of T2-weighted MR imaging and transvaginal sonography. The false-positive diagnoses based on transvaginal sonograms and T2-weighted images, respectively, involved polypoid tumors (n = 4 and 2), distension of the endometrial cavity by pyometra (n = 2 and 1), the presence of myoma (n = 2 and 1), atrophy of the myometrium (n = 1 and 0), and poor tumor/myometrium contrast (n = 0 and 2). On contrast-enhanced MR images, accuracy was influenced only in a case of polypoid tumor, because tumor, endometrial cavity, and myometrium were clearly distinguished and residual myometrium was clearly visualized. With all imaging techniques, false-negative diagnoses were caused mainly by tumors with superficially spreading growth or microscopic invasion. With transvaginal sonography, infiltrative tumor also tended to be understaged (n = 3). Contrast-enhanced MR imaging is significantly superior to transvaginal sonography and unenhanced T2-weighted MR imaging for detecting myometrial invasion.