Objectives To assess the diagnostic accuracy of conventional MRI for detecting and grading subscapularis (SSC) tears by applying the Yoo and Rhee classification. Materials and methods A total of 179 patients who underwent MRI followed by arthroscopic rotator cuff surgery were enrolled. Two musculoskeletal radiologists evaluated the SSC using axial, oblique sagittal, and oblique coronal MRI according to the Yoo and Rhee classification. Using arthroscopic findings as the reference standard, the sensitivity, specificity, accuracy, and interobserver agreement of SSC tears were analyzed. Results Arthroscopy confirmed that the numbers of type I, IIA, IIB, III, IV, and V tears were 35, 70, 35, 9, 9, and 0, respectively. The sensitivity, specificity, and accuracy of readers 1 and 2 for the detection of tears (type IIA or higher) were 85%, 75%, and 82%, and 89%, 70%, and 83%, respectively, while those for the detection of surgical candidates (type IIB or higher) were 77%, 75%, and 75%, and 77%, 83%, and 82%, respectively. The interobserver agreement for detecting SSC tear presence was substantial (kappa = 0.70) for reader 1 vs. reader 2, and those for detecting the surgical candidate group was substantial (kappa = 0.68) for reader 1 vs. reader 2. The interobserver agreement for grading SSC tears was excellent (kappa = 0.86) for reader 1 vs. reader 2. Conclusion Conventional MRI showed 82.5% and 78.5% average accuracy in detecting IIA and IIB or higher tears by applying the Yoo and Rhee classification for the diagnosis of SSC tears with an excellent interobserver agreement in tear grading.
