A prospective comparison of computerized arthrotomography and magnetic resonance imaging of the glenohumeral joint.

Twenty-five patients with shoulder instability or shoulder pain of undetermined etiology were prospectively evaluated with magnetic resonance imaging and computerized arthrotomography. Actual lesions were determined by arthroscopy or at the time of open surgical repair. The images obtained were interpreted independently by three radiologists blinded to both surgical results and the results of previous diagnostic tests. Sensitivity, specificity, and accuracy were determined for each imaging technique for a variety of pathologic entities, including anterior and posterior labral abnormalities, capsular redundancy, biceps-labral complex abnormalities, humeral head (Hill-Sachs) impression lesions, and glenohumeral loose bodies. Analysis of imaging techniques also included construction of receiver operator curves for labral abnormalities. Magnetic resonance imaging showed better diagnostic results in the evaluation of glenoid labral and humeral head impression lesions (P < 0.05). Both imaging techniques were equally successful in identifying biceps-labral lesions and intraarticular loose bodies within the glenohumeral joint. Neither imaging technique was consistent in the evaluation of capsular redundancy. Receiver operator curve analysis confirmed that magnetic resonance imaging was the more accurate imaging study in evaluating anterior and posterior glenoid labral abnormalities.