AIM: To determine the diagnostic accuracy of CT-colonography (CTC) for colorectal nonpolypoid lesions. MATERIALS AND METHODS. In the period 2010-2011, 51 out of 454 patients undergoing CTC received also optical colonoscopy (OC). Three human readers with high, intermediate and low expertise interpreted the images. Flat lesions were defined as 3 mm or less in height; laterally spreading type (LST) lesions were defined as nonpolypoid lesions with more than 10 mm lateral diameter. RESULTS: A total of 75 nonpolypoid colorectal lesions were identified in 21 patients: 43 type II-A low-grade adenomas, 2 type II-c Tis adenocarcinomas, 2 LST Tis adenocarcinomas, 24 nonadenomatous (hyperplastic) lesions and 4 LST infiltrating tumors (T1N0M0 in 2 cases and T2N0M0 in 2 cases). Per-lesion sensitivity and NPV were 44% and 80.5%, while per-patient sensitivity, specificity, PPV, NPV and accuracy were 80.9%, 93.7%, 89.5%, 88.2%, 88.7%. The readers with high and intermediate experience yielded significantly better diagnostic performances than reader with low experience (p =0.072 and p=0.030). All the infiltrating carcinomas and 75% of tis carcinomas were detected by all the readers. CONCLUSION. CTC showed a low per-lesion and an high per-patient diagnostic accuracy for all nonpolypoid colorectal lesions, but an high ability to detect nonpolypoid colorectal carcinomas. Diagnostic performances are strictly related to the reader experience.