Objective To assess the diagnostic accuracy of preoperative endoscopic ultrasonography (EUS) for tumor size and invasion of non-ampullary duodenal neuroendocrine tumors (NA-DETs) and to compare the efficacy and safety of endoscopic submucosal dissection (ESD) and modified ESD for the treatment of NA-DETs. Methods Data of 22 patients with 22 NAD-NETs confirmed by histopathological examinations from January 2007 to January 2018 were retrospectively analyzed. ESD was performed on 13 tumors, and modified ESD was performed on 9 tumors. R0 resection rate, procedure time and incidence of procedure-related complications in the ESD group and the modified ESD group were compared. The postoperative pathological results were used as the gold standard to assess the accuracy of preoperative EUS in diagnosing tumor size and invasion of NA-DETs. Results The mean size of NA-DETs was 6.9±1.5 mm. The accuracy in assessing the invasion depth by EUS was 95.5% (21/22) compared with histological results. R0 resection was achieved in 13/13 (100.0%) of the ESD group and in 7/9 (77.8%) of the modified ESD group (P=1.000). The procedure time was significantly shorter in the modified ESD group than that in the ESD group（16.0±2.2 min VS 29.8±4.9 min，P<0.001）. Intraoperative perforation occurred in one patient and delayed perforation occurred in one patient in the ESD group. Delayed bleeding occurred in one patient in the modified ESD group. Follow-up data were available in all cases with a mean period of 30.0±24.8 months. No cases of local recurrence or distant metastasis were detected in the follow-up period. Conclusion EUS can accurately assess the size and depth of NAD-NETs. Modified ESD can provide comparable clinical outcomes to ESD for NAD-NETs ≤10 mm in diameter that are confined to the submucosa.