A Simple O-RADS-Based Ultrasound Scoring System for Risk Stratification of Ovarian Cystic Lesions: A Multicenter Study.

OBJECTIVE: To develop and validate a simple and practical scoring system based on O-RADS ultrasound features for distinguishing benign from malignant ovarian cystic lesions. METHODS: This multi-institutional retrospective study included women with ovarian cystic lesions who underwent pelvic ultrasound and had either histopathological confirmation or at least 2 years of imaging follow-up. Participants were randomized into training (70%, n=1715) and internal validation (30%, n=711) sets, with an external validation set (n=278) from another center. Univariate and multivariate logistic regression identified independent O-RADS predictors of malignancy, which were assigned weights to construct a scoring system. Performance was validated internally and externally, and compared with O-RADS. Inter-observer agreement was assessed between two junior sonographers and two senior sonographers. RESULTS: Seven independent predictors were identified: maximum diameter ≥10 cm (score 2), solid component (2), irregular internal wall (2), color score 3-4 (2), ascites (2), >3 papillary projections (1), and acoustic shadowing (-2), with an optimal cutoff value of 4 points. In the training set, the scoring system achieved an AUC of 0.973, sensitivity of 95.3%, and specificity of 92.5%, compared to O-RADS (AUC of 0.954, sensitivity of 97.3%, specificity of 86.7%). The scoring system maintained stable performance in internal validation (AUC of 0.968, sensitivity of 93.8%, specificity of 91.8%) and external validation (AUC of 0.964, sensitivity of 93.0%, specificity of 90.2%). Inter-observer agreement was excellent among all sonographers (κ=0.80, p<0.001), as well as between junior (κ=0.82) and senior (κ=0.79) sonographers. CONCLUSION: The O-RADS-based scoring system demonstrated excellent and stable diagnostic performance, potentially offering a simple and practical tool for clinical practice; however, prospective multicenter validation in diverse populations is warranted before widespread adoption.