Abstract
Objective:
Accurate characterization of adnexal masses is essential for determining malignancy risk and guiding clinical management. While the O-RADS MRI system provides standardized risk stratification, the impact of MRI acquisition protocols on diagnostic performance and interrater reliability remains unclear. This study aimed to evaluate the diagnostic performance of O-RADS MRI, assess interrater reliability between readers with different experience levels, and investigate the contribution of dynamic contrast-enhanced MRI to reader concordance.
Materials and Methods:
A total of 387 patients with 496 adnexal masses were retrospectively analyzed. Histopathologic diagnosis was obtained through surgery or biopsy, while lesions without histopathology were considered benign if they remained stable or regressed during at least 9 months of imaging follow-up. Two radiologists with different levels of experience independently evaluated all MRI examinations using the O-RADS MRI lexicon. Diagnostic performance was assessed using sensitivity, specificity, and receiver operating characteristic (ROC) analysis with area under the curve (AUC). Interobserver agreement was evaluated using the Cohen κ statistic.
Results:
Both readers demonstrated high diagnostic performance. Sensitivity, specificity, and accuracy were 96.8%, 94.1%, and 95.2% for reader 1, and 96.8%, 92.9%, and 94.4% for reader 2. AUC values were 0.976 (95% CI: 0.940-0.992) and 0.970 (95% CI: 0.933-0.989), respectively. Interobserver agreement was substantial for benign–malignant classification (κ = 0.757; 95% CI: 0.624-0.889) and excellent for O-RADS category assignment (κ = 0.941; 95% CI: 0.919-0.962). Diagnostic performance was comparable between dynamic and nondynamic protocols; however, dynamic MRI demonstrated higher interobserver agreement, particularly for lesion classification.
Conclusion:
O-RADS MRI provides high diagnostic performance and strong interobserver agreement in the evaluation of adnexal masses. Although dynamic contrast-enhanced MRI does not significantly improve diagnostic accuracy, it may enhance reader concordance, particularly in lesions with equivocal imaging features.