The aims of this prospective study were 1) to evaluate the accuracy of high-resolution three-dimensional endorectal ultrasonography in distinguishing slight from massive submucosal invasion of early rectal tumors, and 2) to determine the technology's role in treatment selection. A total of 142 consecutive patients with clinically possible pT1 rectal cancers underwent three-dimensional endorectal ultrasonography. Slight or massive irregularity of the hyperechoic submucosal layer was considered to characterize uT1-slight or uT1-massive tumors. Treatment was selected on the basis of ultrasonographic findings: endoscopic resection or full-thickness transanal local excision was selected for uT1-slight lesions, and radical resection was selected for uT1-massive tumors. Ultrasonographic staging was compared with histopathologic staging. One hundred twenty-six patients were included in the final analyses. Three-dimensional endorectal ultrasonography staged 77 lesions as uT0, 25 as uT1-slight, 20 as uT1-massive, and 4 as uT2. Histologically, adenomas were found in 75 patients and tumor invasion was found in 44 lesions (24 pT1-slight, 16 pT1-massive, 4 pT2). The overall kappa for the concordance between ultrasonographic and histopathologic stagings was 0.81 (95% confidence interval, 0.72-0.89). No invasive carcinomas remained undetected. The depth of invasion was correctly determined in 87.2% of both pT1-slight and pT1-massive lesions. Considering the complete series of 126 patients, the accuracy of this modality in selecting appropriate management was 95.2% (kappa, 0.84; 95% confidence interval, 0.71-0.96). Adequate surgery was performed in 87.5% of pT1 tumors. Three-dimensional endorectal ultrasonography is useful for assessing the depth of submucosal invasion in early rectal cancer and for selecting therapeutic options.