Accuracy of routine MRI in lesions of the supraspinatus tendon -: Comparison with surgical findings

Purpose: To determine the correlation of the extent of lesions of the supraspinatus tendon in MRI's of the shoulder with surgical or arthroscopic findings using the classification of Snyders and Batemann, respectively. Materials and Methods: The preoperative MRI's of 80 patients (age: 16-76/47.4+/-14.0 years) which were performed due to various complaints of the shoulder were analyzed retrospectively by two experienced and blinded radiologists. We evaluated the incidence and the extent of partial or complete ruptures of the supraspinatus tendon. After MRI, an arthroscopic or open surgical intervention was performed (= gold standard). Various MR-scanners were used with a field strength of 1.0 T (17 cases), or 1.5 T (63 cases) and flexible or inflexible arthro coils. Additionally to plain MRI, 38 of 80 patients underwent contrast enhanced MRI. The MR and the surgical or arthroscopic findings were compared and statistically analyzed. Results: Neglecting the localization and the extent of discontinuity of the supraspinatus tendon, the sensitivity of the 80 MRI's was 0.93, the specificity 0.69, and the accuracy 0.85. The sensitivity increased to 0.96 with constant specificity and an accuracy of 0.83 excluding the lesions with an extent below 1 cm. Compared with non-enhanced examinations, the contrast enhanced MRI revealed higher sensitivity (+ 7%, 0.89 vs. 0.96), higher specificity (+ 11 %, 0.64 vs. 0.75), and higher accuracy (+ 8 %, 0.81 vs. 0.89) in depicting lesions of the supraspinatus tendon at all. On T,weighted images, the detection of lesions at all and the differentiation between partial and complete ruptures were improved significantly by contrast enhancement, especially in lesions with an extent below 1 cm. Diagnostic failures were seen in examinations without intravenous contrast application, artifacts, extent of the lesion below 1 cm, differentiating between degeneration and partial rupture of the tendon, differentiating between severely degenerated tendons with partial rupture and complete rupture, and covered ruptures. Conclusions: The performance and consecutively the analysis of the widely used MRI of the shoulder is often not sufficiently reliable in its present routine form. To be able to compete with other imaging modalities such as sonography, standardized MR protocol, contrast enhancement as needed, avoiding the described source of failure, and an exact analysis should be applied.