Abstract:
The aim of this retrospective study was to determine the prevalence of ovarian masses and calculate the diagnostic performance of the pattern recognition approach in ovarian pathology. A total of 1001 patients diagnosed with ovarian mass were included, of which 92.6% were diagnosed with ovarian pathology and the presence of a pathological result, while 7.4% of cases were diagnosed with functional ovarian cyst. The prevalence of ovarian malignancy was 15%. A specific ultrasound diagnosis was suggested in 62.9% of all cases, while sonographers did not explicitly provide a diagnosis in remaining cases. A subjective assessment showed 80.3% sensitivity (95% confidence interval (CI) 68.7-89.1) and 97.6% specificity (95% CI 96-98.6) in differentiating between benign and malignant ovarian masses. The sensitivity and specificity for the diagnosis of endometriotic cyst were 77.03% and 90.63% and 63.19% and 94.3% for mature cystic teratoma, respectively. In conclusion, assessment showed good performance in differentiating between benign and malignant ovarian mass and it was possible to diagnose several specific ovarian tumours. Impact StatementWhat is already known on this subject? Pattern recognition is an acceptable method for classifying ovarian mass, which exhibits specific morphological features on grey-scale ultrasonography, and can be used to predict nature and histological type.What do the results of this study add? Even in the hands of an expert examiner, there were a number of cases in which the diagnoses could not be specifically stated. Pattern recognition correctly classified 90.3% of ovarian masses as either benign or malignant and correctly provided specific histologic diagnoses after exclusion of unspecified diagnosis in 80.6% of all cases. The diagnostic performance of this approach was high in differentiating between benign and malignant ovarian mass and in diagnosing some specific ovarian pathologies.What are the implications of these findings for clinical practice and/or further research? A subjective assessment is simple and easy to use in clinical practice and has shown promising results in classifying benign and malignant ovarian mass. Moreover, it can also be used to make some specific diagnoses. However, specialised and experienced gynaecological ultrasound examiners are required to provide the most accurate diagnosis. Therefore, criteria to describe ultrasound features and convincing operators to make a definite diagnosis as often as possible should be encouraged. A prospective study to verify diagnostic performance of pattern recognition or comparing with other ultrasonographic diagnostic tools should be considered.
摘要:
这项回顾性研究的目的是确定卵巢肿块的患病率,并计算模式识别方法在卵巢病理学中的诊断性能。共纳入1001例诊断为卵巢肿块的患者,其中92.6%被诊断为卵巢病理并存在病理结果,而7.4%的病例诊断为功能性卵巢囊肿。卵巢恶性肿瘤的患病率为15%。在所有病例中,62.9%建议进行特定的超声诊断,而超声医师没有明确提供其余病例的诊断。主观评估显示,在区分良性和恶性卵巢肿块方面,敏感性为80.3%(95%置信区间(CI)68.7-89.1)和特异性为97.6%(95%CI96-98.6)。对成熟囊性畸胎瘤诊断子宫内膜异位囊肿的敏感性和特异性分别为77.03%和90.63%和63.19%和94.3%,分别。总之,评估显示在区分良性和恶性卵巢肿块方面表现良好,并且可以诊断几种特定的卵巢肿瘤。模式识别是一种可接受的卵巢肿块分类方法,在灰度超声检查中表现出特定的形态学特征,并可用于预测性质和组织学类型。这项研究的结果补充了什么?即使在专家考官手中,有一些病例无法具体说明诊断.模式识别将90.3%的卵巢肿块正确分类为良性或恶性,并在所有病例的80.6%中排除未指明的诊断后正确提供了特定的组织学诊断。这种方法的诊断性能在区分良性和恶性卵巢肿块以及诊断某些特定的卵巢病变方面很高。这些发现对临床实践和/或进一步研究有什么意义?主观评估在临床实践中简单易行,并且在对良性和恶性卵巢肿块进行分类方面显示出了有希望的结果。此外,它也可以用来做一些具体的诊断。然而,需要专业和有经验的妇科超声检查者来提供最准确的诊断。因此,应鼓励描述超声特征的标准,并说服操作员尽可能多地做出明确的诊断。应考虑进行前瞻性研究,以验证模式识别的诊断性能或与其他超声诊断工具进行比较。