The diagnosis of SLAP lesions is a clinical and radiological challenge due to the heterogenous clinical presentation of affected patients as well as the complexity and size of the SLAP complex. MRA is currently regarded as the gold standard of imaging for the detection of SLAP lesions. However, the vast majority of MRI examinations of the shoulder are performed as conventional, non-contrast enhanced MR exams . Additionally, most shoulder MRI exams are conducted in outpatient facilities and interpreted by radiologists without fellowship training in musculoskeletal imaging, contributing to a lower sensitivity for the detection of SLAP lesions . Therefore, improvements in conventional imaging protocols to facilitate diagnosis are warranted. In our study, we evaluated a 3D-MEDIC sequence for its ability to detect SLAP lesions of the shoulder and compared its performance to 2D-PD fs of conventional imaging protocols. The results were correlated with those of subsequent shoulder arthroscopy as a gold standard. Compared to shoulder arthroscopy, 3D-MEDIC imaging yielded an excellent sensitivity of 100% and an excellent NPV of 100% for the detection of SLAP lesions for both experienced and less experienced readers. With regards to the less experienced reader, sensitivity of 3D-MEDIC was even higher than the sensitivity of conventional 2D-PD fs. In other words, 3D-MEDIC imaging was able to detect all lesions and excluded SLAP lesions regardless of the experience of the reading radiologist. In a single case, a SLAP lesion confirmed subsequently was only detectable on 3D-MEDIC but not on 2D-PD fs sequences, contributing to the increased visibility. However, this occurred at a cost of moderate to high-moderate specificity with three false positive diagnoses for the less experienced reader and two false positive diagnoses for the experienced reader. The combination of 3D-MEDIC and 2D-PD fs sequences of the conventional protocol proved to be beneficial. Compared to the diagnostic performance of 3D-MEDIC alone, the diagnostic performance of combined 3D-MEDIC and 2D-PD fs led to an increase in specificity for both readers. Moreover, the experienced reader reached excellent performance values (sensitivity, specificity, PPV, and NPV) using the combined methods. Several possibilities may explain the excellent sensitivity of 3D-MEDIC. MEDIC is a T2* weighted gradient-echo sequence specifically designed for musculoskeletal and neuroradiological purposes and combines up to six echoes in a single image leading to a higher signal-to-noise ratio and reduced susceptibility [13, 15, 18]. Compared to other 3D sequences at the wrist, 3D-MEDIC exhibits a high contrast and signal-to-noise ratio as well as the best visibility of fibrocartilaginous and ligamentous tissue . At 7 T a MEDIC sequence exhibited a high level of anatomical detail with regards to the labrum as well as good labrum/fluid contrast at the hip . Furthermore, the continuous 3D slice acquisition down to 1 mm slice thickness leads to reduction of partial volume artefacts, thereby increasing the visibility of smaller lesions. Moreover, the acquisition of a 3D image stack with an isotropic voxel size of 1.0 × 1.0 × 1.0 mm allows high-resolution multiplanar reconstructions in any desired plane to further increase the conspicuity of smaller lesions [12, 20]. One fact that limits routine 3D imaging of the shoulder is the longer acquisition time compared to that of a single 2D sequence combined with lower robustness. Some 3D sequences at the shoulder have acquisition times of up to 9.5 min . The 3D-MEDIC acquisition time in our study was about 5 min and included the entire shoulder joint. The short acquisition time and robustness of the sequence itself may have contributed to the fact that all examinations and reconstructions in the present study were evaluable. As 3D-MEDIC itself has not been systematically evaluated with regards to SLAP lesions, there is a paucity of literature for comparing our findings. Lee et al. paired GRE sequences, namely a 2D-MEDIC sequence and 3D dual echo steady-state (DESS), and compared both sequences against a combination of conventional T1-weighted spin echo (SE) sequences at 1.5 T with respect to glenoid labral tears including superior, anterior, and posterior labral tears . With arthroscopy as a reference, the combination of both GRE sequences showed a higher sensitivity of 88% compared to that of T1 SE combination , in line with our findings. However, the use of a combination of two different GRE sequences, lower field strengths, different assessment, and the use of a 2D instead of a 3D-MEDIC limits the comparability of their findings to our results. Pahwa et al. used a 3D-MEDIC sequence to evaluate the fibroligamentous structures at the wrist and compared the results to those of conventional MRI and MRI arthrography. They reported a higher sensitivity of 3D-MEDIC for detecting tears of the triangulate fibrocartilage complex and ligamentous structures of the wrist compared to that of conventional 2D-PD fs sequences with open surgery or arthroscopy as reference. This is in line with our results, although the comparability is limited by differences in the field strength and joint of interest . Nevertheless, the findings of Pahwa et al. and Lee et al. as well as our own findings support the assumption that 3D-MEDIC sequence has high sensitivity for detecting fibroligamentous pathologies. The diagnostic performance of conventional MRI for the detection of SLAP lesions has been evaluated in various studies and meta-analyses. In two recent meta-analyses, the pooled sensitivity for the detection of SLAP lesions was estimated to be between 63 to 76% [22, 23]. However, in several studies, the sensitivity of conventional MRI was as low as 38 to 46% [10, 22,23,24]. Indeed, Connolly et al. reported a sensitivity as low as 19% for the detection of SLAP lesions by radiologists without fellowship training in musculoskeletal radiology . As stated above, the vast majority of shoulder examinations are performed using a conventional MRI protocol and conducted in outpatient facilities . In relation to the aforementioned published sensitivities concerning SLAP diagnosis in conventional MRI, the need for improvements of conventional MRI protocols is evident. The sensitivity of 3D-MEDIC in our study was higher than the pooled sensitivity published in a meta-analysis of conventional MRI. A careful reevaluation of the false positive cases led to the assumption that severe degenerative or mucoid changes within the labrum with a non-homogenous increase in internal signals of the labrum itself may resemble tears or at least limit diagnostic certainty. In line with this assumption, Loredo et al. evaluated two gradient-echo sequences in a cadaveric study of the glenoid labrum and observed changes of the labral signal intensity which were histopathologically attributed to mucoid or eosinophilic degeneration, calcification, ossification, fibrovascular tissue, synovial tissue or a combination of these . Therefore, 3D-MEDIC seems to be sensitive to changes in the fibrocartilage matrix [25, 26]. For cases with severe degenerative changes of the labrum in particular, the thorough evaluation of 2D-PD fs sequences may add diagnostic confidence. The interreader agreement measured by Cohen’s kappa for the detection of SLAP lesions was 0.82 for the 3D-MEDIC and 0.87 for the PD fs sequences, consistent with an almost perfect agreement according to Koch and Landis . This indicates the reproducibility of the findings using both sequences which is crucial for reliable image interpretation.
There are several limitations to our study. First, our study sample was small, and the results regarding diagnostic performance have to be verified in further (multi-centric) studies with larger patient cohorts. Furthermore, we were unable to perform subgroup analysis according to different subtypes of SLAP lesions due to the small sample size. In addition, the patient cohort was recruited in a tertiary referral center and was therefore highly selected. Hence, the results may not be generalizable to the general population. Moreover, we only focused on the diagnosis of SLAP lesions. It has been described that SLAP lesions are common findings in asymptomatic middle-aged patients . Therefore, future studies are needed to evaluate the clinical impact of an improved detection of SLAP lesions in different settings.
Further, the use of GRE sequences like 3D-MEDIC may be limited in patients with metal implants, due to the increased susceptibility for artefacts . As we excluded patients who undergone shoulder surgery, our findings may only be applicable to patients without prior surgery. This is of great interest as the detection of complications as well as newly acquired shoulder pathologies in a postsurgical setting is crucial for the diagnosis and the initiation of treatment measures . Therefore, further studies examining the feasibility of the 3D-MEDIC in different surgical setting with metallic and non-metallic implants are necessary. Moreover, future studies are required to validate the findings of our study in larger and less selected patient cohorts. Also, studies comparing the performance of 3D MEDIC with MR arthrograms in the same patient cohort are desirable. Future studies may also investigate whether the increase in diagnostic performance leads to a change in patient treatment or patient outcome.