Result: 增强MRI瘤内瘤周影像组学联合临床影像学 特征评估肝细胞癌ki-67的表达.

Objective To construct a model for predicting ki-67 expression in hepatocellular carcinoma using the intratumoral and peritumoral radiomic features of contrast enhanced magnetic resonance imaging （CEMRI） in the arterial phase as well as clinical imaging features. Methods A total of 120 patients pathologically diagnosed with hepatocellular carcinoma （HCC） from January 2016 to December 2024 in No. 910 Hospital of the Joint Logistics Support Force of the Chinese People's Liberation Army were retrospectively enrolled and randomly divided into a training set （84 cases） and a test set （36 cases） in a ratio of 7∶3. ITK-SNAP software was used to delineate the global region of interest （ROI） of HCC on the arterial phase MR images. The ROIs of all patients were automatically expanded outward by 2 mm， and then the intratumoral ROI areas were eliminated to obtain the peritumoral ROI. With the help of PyRadiomics software， 1 198 intratumoral and peritumoral radiomic features were extracted. Spearman correlation analysis， maximum relevance-minimum redundancy （mRMR）， and least absolute shrinkage and selection operator （LASSO） regression were used to reduce the data dimension and select the best features. Then， a radiomics model of the logistic regression （LR） machine learning algorithm was constructed. A combined model including clinical imaging features and radiomics features was established. The area under the curve （AUC）， accuracy， sensitivity， specificity， positive predictive value （PPV）， negative predictive value （NPV）， calibration curve and decision curve analysis （DCA） were used to evaluate the efficacy of the intratumoral and peritumoral radiomics features combined with clinical imaging features model in predicting ki-67 expression in hepatocellular carcinoma. Results The intratumor model exhibited an efficacy in predicting the expression of ki-67 in hepatocellular carcinoma with AUC values of 0.817 and 0.787 in the training set and test set， respectively. The peritumoral model showed an efficacy with AUC values of 0.805 and 0.633 in the training set and test set， respectively. The intratumoral and peritumoral model demonstrated AUC values of 0.874 and 0.836 in the training set and test set， respectively. The combined model constructed by integrating the intratumoral and peritumoral model with clinical imaging features yielded AUC values of 0.877 and 0.849 in the training set and test set， respectively， indicating clinical imaging features improved the performance of the model. DCA showed that the combined models all had good clinical benefits， with the intratumoral and peritumoral model performing the best. Conclusion The intratumoral and peritumoral radiomics model based on CEMRI arterial phase combined with clinical imaging data can accurately predict the expression of ki-67 in hepatocellular carcinoma. This combined model yields the best clinical benefit. [ABSTRACT FROM AUTHOR]

目的 旨在基于增强磁共振成像（contrast enhanced magnetic resonance imaging， CEMRI）的动 脉期相瘤内及瘤周影像组学特征，结合临床影像学特征构建预测肝细胞癌（hepatocellular carcinoma， HCC）ki-67表达的模型。方法 回顾性纳入中国人民解放军联勤保障部队第910医院2016年1月至2024年 12月经手术切除病理证实为HCC，且符合纳入标准和排除标准的患者120例，按7∶3的比例随机划分为训 练集（n = 84 例）和验证集（n = 36）。使用ITK-SNAP 软件在动脉期相MR 图像上勾画HCC 全域感兴趣区 （region of interest， ROI），并针对所有患者的ROI，自动向外扩张2 mm，随后剔除瘤内及超出肝脏边缘的区 域，以此获取瘤周ROI。利用PyRadiomics 开源程序包在Python 软件分别提取瘤内（intratumoral）、瘤周 （peritumoral）1 198 个影像组学特征，并采用Spearman 相关性分析、最大相关性-最小冗余（maximum relevance-minimum redundancy， mRMR）和最小绝对收缩和选择算子（least absolute shrinkage and selection operator， LASSO）回归进行数据降维并选择其最佳特征，然后构建逻辑回归（Logistic Regression， LR）机 器学习算法的影像组学模型，再结合临床影像学特征，建立一个包含临床影像特征和影像组学特征的组 合模型。使用受试者工作特征（receiver operating characteristic， ROC）曲线下面积（area under the curve， AUC）、精度（accuracy）、灵敏度（sensitivity）、特异性（specificity）、阳性预测值（positive predictive value， PPV）、 阴性预测值（negative predictive value， NPV）、校正曲线（calibration curve）、决策曲线（decision curve analysis， DCA）评估瘤内瘤周影像组学特征联合临床影像学特征的组合模型预测HCC ki-67表达的价值。结果 瘤内 组学模型预测HCC ki-67表达能力的训练集的AUC值为0.817，验证集的AUC值为0.787；瘤周组学模型预 测的AUC值在训练集为0.805，在验证集的AUC值为0.633；瘤内瘤周联合模型预测肝细胞癌ki-67效能最佳， 在训练集和验证集AUC值分别为0.874、0.836。使用瘤内瘤周模型联合临床影像学特征构建组合模型，训 练集和验证集的AUC 值分别为0.877、0.849。DCA 提示组合模型均具有良好的临床收益。结论 基于 CEMRI动脉期相的瘤内瘤周影像组学特征联合临床影像特征的组合模型，能准确预测HCC ki-67的表达， 临床效益最佳. [ABSTRACT FROM AUTHOR]