摘要
目的探讨基于T2WI影像组学方法鉴别肝细胞癌(hepatocellular carcinoma,HCC)和肝内胆管细胞癌(intrahepatic cholangiocarcinoma,ICC)的价值。方法回顾性分析我院经病理证实的63例HCC和23例ICC的患者的临床资料。运用Darwin research platform将63例HCC患者随机分成训练组(50例)和测试组(13例),23例ICC患者随机分成训练组(18例)和测试组(5例),对所有病灶T2WI提取组学特征,构建有效影像组学模型。结果在逻辑回归降维后筛选出7个特征,建立支持向量机和逻辑回归两个有效影像组学模型,逻辑回归模型训练组AUC值为0.925,灵敏度、准确度和特异度分别为0.833、0.911、0.833;测试组AUC值为0.892,灵敏度、准确度和特异度分别为0.800、0.833、0.666;支持向量机模型训练组AUC值为0.928,灵敏度、准确度和特异度分别为0.888、0.882、0.727;测试组AUC值为0.861,灵敏度、准确度和特异度分别为0.800、0.777、0.571。结论基于T2WI图像构建的影像组学模型能有效鉴别HCC和ICC,可协助临床进行诊断,降低误诊率。
Objective To explore the value of method based on T2 WI radiomics in the identification of hepatocarcinoma(HCC)and intrahepatic cholangiocarcinoma(ICC).Methods The clinical datas of sixty-three cases of HCC patients and 23 cases of ICC patients confirmed by pathology in our hospital were retrospectively analyzed.Sixty-three cases of HCC patients were randomly divided into training group(50 cases)and verification group(13 cases),and 23 cases of ICC patients were randomly divided into training group(18 cases)and verification group(5 cases)by Darwin research platform.The group characteristics were extracted from T2 WI of all lesions to construct an effective radiomics model.Results After the dimension reduction of Logistic regression,seven features were screened out,and two effective radiomics models,support vector machine and Logistic regression,were established.AUC value of the Logistic regression model training group was 0.925,sensitivity,accuracy and specificity were 0.833,0.911 and 0.833,respectively;AUC value of the verification group was 0.892,sensitivity,accuracy and specificity were 0.800,0.833 and 0.666,respectively;AUC value of the support vector machine model training group was 0.928,sensitivity,accuracy and specificity were 0.888,0.882 and 0.727,respectively;AUC value of the verification group was 0.861,sensitivity,accuracy and specificity were 0.800,0.777 and 0.571,respectively.Conclusion The radiomics model built based on T2 WI image can effectively differentiate HCC and ICC,which can assist clinical diagnosis and reduce misdiagnosis rate.
作者
冯忠园
叶靖
FENG Zhongyuan;YE Jing(the Hospital of Yizheng People's Affiliated to Yangzhou University Medical College,Yangzhou 211400;Northern Jiangsu People's Hospital,Yangzhou 225002,China)
出处
《临床医学研究与实践》
2020年第22期1-4,共4页
Clinical Research and Practice
