OBJECTIVE To compare the diagnostic accuracy of three-dimensional (3D) fast spin echo (FSE) intermediate-weighed (IW-3D) and 3D hybrid double-echo steady-state T1-weighted sequences (Hy-3D) and two-dimensional (FSE) images (2D) at shoulder MR arthrography (MRA). METHODS Institutional review board approval was obtained and informed consent was waived for this retrospective study. From September 2011 to October 2014, 102 patients who had undergone 1.5 Tesla MRA of the shoulder, including conventional 2D-FSE and IW-3D and Hy-3D images were included in our study. The mean interval between MRA and surgery was 21 days (range 2-70 days). MR images were retrospectively and independently reviewed by two experienced radiologists blinded to the clinical and surgical data. Supraspinatus tendon (SST), infraspinatus tendon (IST) and subscapularis tendon (SCT) tears, as well as antero-inferior, superior and posterior labral lesions were assessed, using surgery as the reference standard. Each reader's performance in assessing rotator cuff and labrum abnormalities was evaluated using the area under the receiver operating characteristic curve (AUC) and 95% confidence intervals (CIs). The difference was evaluated using a univariate z test. The sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV) and accuracy (Acc) for all types of rotator cuff tears and labral lesions were calculated. A value of p < 0.05 was considered statistically significant. Inter-observer agreement was calculated using kappa statistics. RESULTS The difference of diagnostic accuracy achieved was not significant (p > 0.05). In particular, differences in AUC values ranged from 0.002 (p = 0.98) to 0.014 (p = 0.82) as regards the comparison between 2D and IW-3D images, from 0.002 (p = 0.98) to 0.034 (p = 0.08) concerning the comparison between 2D and Hy-3D images and from 0.010 (p = 0.82) to 0.032 (p = 0.09) when comparing Hy-3D to IW-3D images. Accuracy values in evaluating RC lesions and labral lesions were 95.1, 92.1, 91.2, 93.1, 93.1 and 94.1% by reading 2D, Hy-3D and IW-3D images, respectively. The difference of diagnostic accuracy achieved using the datasets analyzed was not significant (p > 0.05). Inter-observer agreement was very good for each of the datasets that were evaluated, with near-perfect agreement for 2D dataset (k = 0.86), Hy-3D (k = 0.81) and IW-3D (k = 0.83). CONCLUSIONS The accuracy of IW-3D and Hy-3D images was not significantly higher than the 2D sequences in evaluating RC and labral lesions.