中国中西医结合影像学杂志2025，Vol.23Issue(6)：677-682,687,7.DOI:10.3969/j.issn.1672-0512.2025.06.002
基于MRI增强扫描图像的影像组学特征与DWI特征联合模型对肝细胞癌患者CK19表达水平的预测
Prediction of a model based on MRI enhancement combined with DWI for CK19 expression status in patients with hepatocellular carcinoma
摘要
Abstract
Objective:To explore the value of MRI enhancement radiomics combined with DWI features to construct a model for predicting CK19 expression level in patients with hepatocellular carcinoma(HCC).Methods:A total of 171 HCC patients were retrospectively analyzed and randomly divided into a training cohort(120 cases)and a validation cohort(51 cases)at a ratio of 7∶3.The training cohort was divided into the low expression group(93 cases)and the high expression group(27 cases)according to the results of immunohistochemical CK19 staining after operation,while in the validation cohort,there were 38 cases in the low expression group and 13 cases in the high expression group.The radiomics features were extracted from the arterial phase(AP),portal venous phase(VP),delayed phase(DP)and three-phase images on Radcloud platform.The variance threshold method,SelectKBest,and least absolute shrinkage and selection operator(LASSO)algorithms were used to screen radiomics features.Then four corresponding radiomics prediction models,AP,VP,DP,and AP+VP+DP models were constructed using the support vector machine(SVM)classifier.The DWI prediction model was constructed based on DWI image and ADC values.The model with the best performance in radiomics was combined with the DWI model to construct a combined prediction model,and the effectiveness of the combined model was evaluated.ROC curves were used to evaluate the effectiveness of each model in predicting CK19 expression levels.Results:Among the four radiomics models,the AP+VP+DP model had the best performance,with AUCs of 0.855 and 0.818 in the training and validation cohorts,respectively.In the training cohort,the high expression group had a higher proportion of target ring sign on DWI and the lower ADC value compared with the low expression group,and the differences were statistically significant(both P<0.05).Multiple logistic regression analysis showed that DWI target ring sign and ADC value were independent predictors of CK19 high expression in HCC(both P<0.05).The DWI model was constructed using DWI target ring sign and ADC value,and the AUCs in the training and validation cohorts were 0.765 and 0.696,respectively.The combined model constructed by AP+VP+DP radiomics features,DWI target ring sign and ADC values had the AUCs in the training and validation cohorts of 0.881 and 0.879,respectively.Conclusion:The AP+VP+DP model based on preoperative MRI enhancement images has a high application value in predicting the CK19 high expression in HCC.The combined model based on radiomics and DWI features shows good efficacy and can be used as a reliable method for clinical preoperative non-invasive prediction of CK19 expression status in HCC.关键词
癌,肝细胞/磁共振成像/细胞角蛋白19/影像组学Key words
Carcinoma,hepatocellular/Magnetic resonance imaging/Cytokeratin 19/Radiomics引用本文复制引用
叶卫川,陈炜越,高瑞杰,王正,吴樟扬,张莹,罗项超,卢陈英,纪建松..基于MRI增强扫描图像的影像组学特征与DWI特征联合模型对肝细胞癌患者CK19表达水平的预测[J].中国中西医结合影像学杂志,2025,23(6):677-682,687,7.基金项目
浙江省医药卫生青年创新项目(2023RC115). （2023RC115）
