曾凤霞,刘仁懿,曾辉,陈卫国,秦耿耿.临床、CT影像组学及融合模型预测肝细胞癌分化程度[J].中国医学影像技术,2021,37(7):1029~1033
临床、CT影像组学及融合模型预测肝细胞癌分化程度
Clinical, CT radiomics and combining models for predicting differentiation degrees of hepatocellular carcinoma
投稿时间:2020-09-10  修订日期:2021-05-02
DOI:10.13929/j.issn.1003-3289.2021.07.015
中文关键词:  癌,肝细胞  细胞分化  体层摄影术,X线计算机  影像组学
英文关键词:carcinoma, hepatocellular  cell differentiation  tomography, X-ray computed  radiomics
基金项目:国家重点研发计划(2019YFC0121903、2019YFC0117301)。
作者单位E-mail
曾凤霞 南方医科大学南方医院放射科, 广东 广州 510515  
刘仁懿 南方医科大学南方医院放射科, 广东 广州 510515  
曾辉 南方医科大学南方医院放射科, 广东 广州 510515  
陈卫国 南方医科大学南方医院放射科, 广东 广州 510515  
秦耿耿 南方医科大学南方医院放射科, 广东 广州 510515 zealotq@smu.edu.cn 
摘要点击次数: 306
全文下载次数: 173
中文摘要:
      目的 评价临床、CT影像组学及融合模型预测肝细胞癌(HCC)分化程度的可行性。方法 纳入330例HCC患者,根据病理所见分化程度分为高分化组(n=85)、中分化组(n=161)及低分化组(n=84),比较组间临床资料及CT征象差异。按3∶1比例随机将各组分为训练集及测试集。提取训练集CT影像组学特征,构建临床模型、影像组学模型及融合模型。绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),评估各模型鉴别不同分化程度HCC的效能。结果 共纳入352个CT影像组学特征,109个来自高、中分化HCC,84个来自中、低分化HCC,159个来自高、低分化HCC。临床模型鉴别高、低分化HCC的AUC为0.85;CT影像组学模型鉴别高分化与中、低分化HCC的AUC分别为0.80及0.79;融合模型鉴别高、低分化HCC的AUC为0.88。结论 临床、CT影像组学及融合模型预测高、低分化HCC的效能均较高。CT影像组学模型可较好地预测高、中分化HCC。
英文摘要:
      Objective To observe the effectiveness of models established according to clinical data, CT imaging radiomics and combining both for predicting differentiation degrees of hepatocellular carcinoma (HCC). Methods A total of 330 HCC patients were enrolled and divided into well-differentiated HCC group (n=85), moderately-differentiated HCC group (n=161) and poorly-differentiated HCC group (n=84). The clinical data and CT signatures were compared among groups. Then patients in each group were randomly divided into training set and test set according at the ratio of 3 ∶1. CT radiomics features of HCCs were selected from training set, and clinical, radiomics and combining models were constructed, respectively. Receiver operating characteristic (ROC) curves were drawn, and the areas under the curve (AUC) were calculated to evaluate the effectiveness of the models in identifying HCC with different differentiation degrees in the test set. Results A total of 352 radiomics features were selected, among which 109 features were selected from well- and moderately-differentiated groups, 84 from moderately- and poorly-differentiated groups, and 159 from well- and poorly-differentiated groups. AUC of the clinical model for differentiating well- and poorly-differentiated HCC was 0.85, of the radiomics model for differentiating well-differentiated and moderately- and poorly-differentiated HCC was 0.80 and 0.79, respectively. AUC of the combining model for differentiating well- and poorly-differentiated HCC was 0.88. Conclusion All of clinical, CT radiomics and combining models high efficacy for predicting well- and poorly-differentiated HCC, while CT radiomics model had good effectiveness for predicting well- and moderately-differentiated HCC.
查看全文  查看/发表评论  下载PDF阅读器