PURPOSE: To evaluate the diagnostic value of MRI of the shoulder. MATERIALS AND METHODS: Between January andJune 1997, shoulder MRI and arthroscopy were performed in a total of 48 patients with shoulder pain (n=30) orshoulder instability (n=18). Forty-five were males and three were females; their ages ranged from 16 to 67 (mean32.5) years. The period between shoulder MRI and arthroscopy was between one and 390 (mean, 42.2) days. Twenty-sixpatents underwent MR arthrography (AMR), and 22 conventional MRI(CMR). Each image was analyzed for rotator cuffinjury, glenoid labral injury, SLAP lesion, and biceps tendon injury. RESULTS: On arthroscopy, one disease wasfound in 34 patients, two were found in six, three diseases were found in seven, and one patient had fourdiseases. Arthroscopic diagnosis was as follows: rotator cuff injury, 29 ; SLAP lesion, 12 ; glenoid labralinjury, 10 ; biceps tendon injury, 4 ; subacromial bursitis, 2 ; chronic synovitis, 1 ; adhesive capsulitis, 1 ;superior glenohumeral ligament injury, 1 ; normal, 1. For rotator cuff injury, the sensitivity of MRI was 65.5%and specificity was 93.0% (AMR : 66.7%, 95.8%, CMR : 65.2%, 86.4%). For SLAP lesion, sensitivity was 58.3% andspecificity was 97.2% (AMR : 66.7%, 100%, CMR : 50%, 93.8%) ; for glenoid labral injury, sensitivity was 80.0% andspecificity was 89.5% (AMR : 85.7%, 84.2%, CMR : 66.7%, 94.7%), and for biceps tendon injury, the false negativerate was 100%. CONCLUSION: In cases involving glenoid labral injury, the diagnostic accuracy of shoulder MRI wasrelatively high ; in rotator cuff injury and SLAP lesion, however, diagnosis was limited, and in biceps tendoninjury was difficult. We suggest, however, that MR arthrography has certain diagnostic advantages overconventional MRI.