Accuracy of transvaginal ultrasound examination for assigning a specific diagnosis to adnexal masses

Objectives: To determine the diagnostic performance of subjective evaluation of ultrasound findings (pattern recognition) with regard to making a specific diagnosis in adnexal masses. Methods: Some 1066 patients in the International Ovarian Tumor Analysis study were included. They underwent transvaginal gray-scale and color Doppler ultrasound examination by an experienced examiner. Using pattern recognition each mass was classified as benign or malignant, and if possible a specific diagnosis was suggested endometrioma, dermoid cyst, hydrosalpinx, functional cyst, paraovarian cyst, tubo-ovarian abscess, serous or mucinous cyst, adenofibroma, torsion, simple cyst, peritoneal cyst, fibroma/thecoma, rare benign tumor, borderline tumor, primary invasive cancer, metastatic cancer or rare malignant tumor. The suggested diagnosis was compared with the histological diagnosis of the surgical specimen. Results: There were 800 benign and 266 malignant tumors. A specific diagnosis was suggested in 900 (84%) of 1066 tumors. Pattern recognition diagnosed dermoid cyst with sensitivity 86% (100/116), specificity 99% (938/950), positive likelihood ratio (LR+) 86, and negative likelihood ratio (LR−) 0.1; endometrioma with sensitivity 77% (153/199), specificity 98% (847/867), LR+ 36 and LR− 0.2; hydrosalpinx with sensitivity 86% (18/21), specificity 98% (1022/1045), LR+ 43 and LR− 0.1. Pattern recognition was associated with high LR+ (8 56) and high LR− (0.3 0.9) for all other diagnoses except peritoneal inclusion cyst (excluding rare benign tumors and simple cysts which were associated with very low LR+ and high LR−). LR+ and LR− for peritoneal inclusion cyst (n = 5) were 80 and 0.2. Conclusions: Using pattern recognition it is possible to make a virtually conclusive diagnosis of dermoid cyst, endometrioma and hydrosalpinx. Many other adnexal pathologies can be recognized, but not confidently excluded, using pattern recognition.