METHODS: This single-center, retrospective study analyzed ultrasound data from January 2017 to March 2020 of 206 women with adnexal masses. Patients who underwent ultrasonography and serum CA-125 measurement and received postsurgery histological diagnosis were included. The ADNEX model\'s diagnostic performance was evaluated by two trainees and two specialists using the area under the receiver operating characteristic curve (AUC) and measures of accuracy, sensitivity, specificity, and predictive values for overall performance and each examiner.
RESULTS: Of the 206 included Japanese women, the prevalence of malignancy was 30.1%, including borderline cases. The overall AUC for distinguishing malignancy was 0.848 (95% confidence interval [CI]: 0.817-0.880). The AUC for each examiner ranged from 0.791 to 0.898, with Specialist 2 showing the highest accuracy and sensitivity varying between 0.677 and 0.839. A moderate degree of agreement was noted among the four examiners (Fleiss\' kappa was 0.586). The performance of trainees and specialists differed significantly in evaluating the solid tissue and the papillary projections in both malignant and benign groups (P < .001).
CONCLUSIONS: The IOTA ADNEX model effectively differentiates benign and malignant adnexal masses in Japanese women. Although the accuracy matched up moderately among the four examiners, better accuracy is expected with training in evaluating solid tissue and papillary projections.
方法:这种单中心,回顾性研究分析了2017年1月至2020年3月206例附件肿块女性的超声数据.包括接受超声检查和血清CA-125测量并接受术后组织学诊断的患者。ADNEX模型的诊断性能由两名受训人员和两名专家使用受试者工作特性曲线下面积(AUC)和准确度测量进行评估。灵敏度,特异性,以及整体表现和每位考官的预测值。
结果:在206名日本女性中,恶性肿瘤的患病率为30.1%,包括边缘案件。区分恶性肿瘤的总体AUC为0.848(95%置信区间[CI]:0.817-0.880)。每位检查者的AUC范围为0.791至0.898,其中专家2显示最高的准确性和灵敏度在0.677至0.839之间变化。四名审查员之间的协议程度适中(Fleiss\'kappa为0.586)。受训者和专家在评估恶性和良性组中的实体组织和乳头状突起方面的表现存在显着差异(P<.001)。
结论:IOTAADNEX模型可有效区分日本女性的良性和恶性附件肿块。尽管四位考官的准确性适度匹配,通过评估实体组织和乳头状投影的训练,预期有更好的准确性。