Diagnostic Performance Evaluation of a Computer-Assisted Imaging Analysis System for Ultrasound Risk Stratification of Thyroid Nodules
Diagnostic Performance Evaluation of a Computer-Assisted Imaging Analysis System for Ultrasound Risk Stratification of Thyroid Nodules
复制标题
DOI：
10.2214/ajr.18.20740
复制
发表时间：
2019-07-01
影响因子：
5
通讯作者：
Puig-Domingo, Manuel
中科院分区：
文献类型：
--
作者：
Reverter, Jordi L.;Vazquez, Federico;Puig-Domingo, Manuel
目标。基于超声的甲状腺结节恶性风险分层具有潜在的可变性。本研究的目的是评估第一个商用的计算机辅助诊断(CADx)成像分析系统的诊断效果。一位甲状腺专家回顾了手术治疗前获得的300个甲状腺结节(其中135个是恶性的)的超声图像，并将他对每个图像的分类与图像分析程序(AmCAD-UT，AmCAD BIOMED)所提供的分类进行了比较。美国甲状腺协会(ATA)分类系统、欧洲甲状腺影像报告和数据系统(EU-TIRADS)以及美国和意大利临床内分泌学家协会(美国临床内分泌学家协会[AACE]、美国内分泌学会[ACE]和美第奇内分泌学协会[AME])联合提出的分类系统用于风险分层。结果：使用ATA系统时，甲状腺专家的诊断表现如下：敏感性87.0%；特异性91.2%；阳性预测值90.5%；阴性预测值90.9%。与专家相比，CADx程序与三种分类系统一起使用时具有相似的敏感性，但特异性和阳性预测值较低。在阴性预测值方面，专家的结果与应用ATA分类系统的CADx程序的结果没有差异(90.9%比86.3%；p=0.07)。当使用ATA分类系统时，专家临床医生的ROC AUC值为0.88，CADx程序的ROC AUC值为0.72。结论：本研究中描述的CAM超声图像分析程序对甲状腺结节的风险分层是有用的，但它的表现不如超声专家。
OBJECTIVE. Ultrasound-based stratification of the malignancy risk of thyroid nodules has potential variability. The purpose of this study is to evaluate the diagnostic effectiveness of the first commercially available system for computer-aided diagnosis (CADx) imaging analysis.MATERIALS AND METHODS. Ultrasound images of 300 thyroid nodules (135 of which were malignant) acquired before surgical treatment were retrospectively reviewed by a thyroid expert, and his classification of each image was then compared with the classification rendered by an image analysis program (AmCAD-UT, AmCAD Biomed). The American Thyroid Association (ATA) classification system, the European Thyroid Imaging Reporting and Data System (EU-TIRADS), and the classification system jointly proposed by American and Italian associations of clinical endocrinologists (the American Association of Clinical Endocrinologists [AACE], the American College of Endocrinology [ACE], and Associazione Medici Endocrinologi [AME]) were used for risk stratification.RESULTS. The diagnostic performance of the thyroid expert when the ATA system was used was as follows: sensitivity, 87.0%; specificity, 91.2%; positive predictive value, 90.5%; and negative predictive value, 90.9%. Compared with the expert, the CADx program, when used with the three classification systems, had a similar sensitivity but a lower specificity and positive predictive value. Regarding the negative predictive value, the results of the expert did not differ from those of the CADx program when it applied the ATA classification system (90.9% vs 86.3%; p = 0.07). The ROC AUC value was 0.88 for the expert clinician and 0.72 for the CADx program when the ATA classification system was used.CONCLUSION. The CAM ultrasound image analysis program described in the present study is useful for risk stratification of thyroid nodules, but it does not perform better than a sonography expert.