Assessment of pattern recognition for the ultrasound diagnosis of an endometrioma.

To describe the US characteristics of masses that, based on pattern recognition, are presumed to be an endometrioma. Methods The IOTA study included 3513 patients with an adnexal mass. All patients were scanned following the same US protocol. The US examiner had to suggest a specific histological diagnosis and to report with which level of confidence he made the diagnosis of benign or malignant. Masses with a presumed diagnosis of endometrioma were included. We considered the following three types of masses as typical for endometrioma: unilocular mass with ground glass appearance, unilocular-solid mass, ground glass and a papillary projection without internal flow, and a multilocular mass with ground glass echogenicitiy. Results Out of 660 masses that were presumed to be endometriosis by the expert, 586 (88.8%) were proven endometriomas. 460 (78%) of them had typical features. Of the 74 misdiagnoses based on subjective impression, 64 (9.6%) proved to be benign and 10 (1.5%) malignant. Of the benign misdiagnoses, 31 (48%) had features typical for endometriomas. Histology was functional cyst (24), abscess (5), hydrosalpinx (5), dermoid (5), serous or mucinous cystadenoma (19), and other benign mass (6). Of the malignant misdiagnoses, 7 (70%) had US features typical for endometriomas and 3 were multilocular-solid tumors. 7 out of the 10 malignant masses were associated with normal CA 125 levels. Histologies were 3 endometrioid adenoca (2 stage I, 1 stage III), 1 stage III serous papillary adenoca, 5 stage I borderline tumors, and one uterine clear cell carcinoma. Sonologists were more confident about their diagnosis in the group that proved to be endometriomas (very confident 82%) than in the group with benign (very confident 59%) or malignant (very confident 10%) misdiagnoses. Conclusions Expert sonolgists can reliably diagnose endometriomas. However, even in “typical” unilocular cysts with ground glass appearance, the rate of unexpected malignancy is 1%.