摘要
目的探讨计算机辅助诊断系统(CADx)在微钙化检测与特征提取基础上的分类对于导管原位癌(DCIS)的诊断价值。方法回顾性分析经南方医科大学附属南海医院及中山大学肿瘤防治中心行乳腺X线摄影检查发现微钙化并经病理学证实的623例患者影像资料,其中,良性病变378例,DCIS 245例。用受试者操作特征曲线(ROC)分别分析采用计算机方法提取的每个微钙化特征对于这两类病变判别的诊断效能,和应用所有微钙化特征集合并基于支持向量机(SVM)分类器的CADx的分类诊断效能。结果 CADx对于良性病变和DCIS这两类病变微钙化分类的ROC曲线下面积(Az)为0.853;特异度、准确率、敏感度分别为70.1%、82.1%、90.7%,高于单个微钙化特征的诊断效能。结论采用CADx对于DCIS微钙化能较好的检测与定位,对乳腺癌早期病变的识别能提供有益的参考。
Objective To explore the classification performance of Computer-aided Diagnosis System（ CADx） for DCIS（ Ductal Carcinoma in Situ） based on the detection and extraction of clustered microcalcifications. Methods Mamammograms from 623 female patients,including 375 benign cases and 248 DCIS,were pathologically confirmed and retrospectively analyzed. ROC curve was used to evaluate the diagnostic performance of each feature of microcalcifications,and the performance of CADx,which employed all features as the input data to SVM classifier. Results For the performance of CADx,area under the ROC curve was 0. 853,with specificity 70. 1%,accuracy 82. 1% and sensitivity 90. 7%. With the combination of all features,the performance of CADx exceeded every single feature. Conclusion CADx in our study gets a good detection and recognition of microcalcifications of DCIS. It provides a promising reference for the earlier diagnosis of breast cancer.
出处
《临床放射学杂志》
CSCD
北大核心
2016年第9期1352-1356,共5页
Journal of Clinical Radiology
基金
佛山市医学类科技攻关项目(编号:2013081750)
广东省科技计划项目(编号:2016B090918066)
作者简介
通讯作者
