首页|期刊导航|分子影像学杂志|肝细胞癌经导管动脉化疗栓塞术治疗反应的预测:基于多期增强CT影像组学联合深度学习

肝细胞癌经导管动脉化疗栓塞术治疗反应的预测:基于多期增强CT影像组学联合深度学习OA

Prediction of transarterial chemoembolization treatment response in hepatocellular carcinoma:multiphasic enhanced CT-derived radiomics combined with deep learning

中文摘要英文摘要

目的 基于术前增强CT动脉期及门脉期图像构建并验证6种预测模型,对比其预测效能并筛选最优模型,以评估肝细胞癌(HCC)患者经导管动脉化疗栓塞术(TACE)治疗反应.方法 纳入175例接受TACE治疗的HCC患者,其中将公开数据集WAW-TACE中的113例患者作为训练集,同时回顾性收集2020年7月~2025年10月齐齐哈尔医学院附属医院中的62例患者作为外部验证集.利用3DSlicer软件在动脉期和门脉期的图像上勾画感兴趣区域(ROI),并从中提取影像组学和深度学习特征.通过单变量分析、相关性分析及最小绝对收缩与选择算子(LASSO)完成特征筛选后,利用支持向量机(SVM)算法构建预测模型.使用曲线下面积(AUC)、准确性、灵敏度和特异性评估模型性能,并在外部验证集中验证预测性能.结果 分别对动脉期和门脉期图像进行特征提取,每期各获得影像组学特征1130个和深度学习特征512个.经筛选与降维后,动脉期联合模型综合性能最优,训练集AUC为0.851(95%CI:0.778~0.924),外部验证集为0.808(95%CI:0.700~0.917),且临床净获益优于单一模型及门脉期联合模型.结论 基于动脉期构建的联合模型可在一定程度上预测HCC患者TACE术后治疗反应,为临床治疗策略制定及预后评估提供辅助参考.

Objective To construct and validate six prediction models based on preoperative enhanced CT images in the arterial and portal venous phases,compare their predictive performance,and identify the optimal model for evaluating treatment response to transarterial chemoembolization(TACE)in patients with hepatocellular carcinoma(HCC).Methods A total of 175 HCC patients who underwent TACE were included.Among them,113 patients from the public WAW-TACE dataset were assigned as the training set,while 62 patients retrospectively collected at the Affiliated Hospital of Qiqihar Medical University from July 2020 to October 2025 served as the external validation set.Regions of interest(ROIs)were manually delineated on arterial phase and portal venous phase images using 3D Slicer software,from which radiomics and deep learning features were subsequently extracted.After feature selection was performed using univariate analysis,correlation analysis,and the least absolute shrinkage and selection operator(LASSO),prediction models were constructed using the support vector machine(SVM)algorithm.Model performance was evaluated using the area under the receiver operating characteristic curve(AUC),accuracy,sensitivity,and specificity,and was validated in the external validation set.Results Feature extraction was performed separately on arterial and portal phase images,yielding 1,130 radiomics features and 512 deep learning features from each phase.After feature selection and dimensionality reduction,the arterial phase combined model performed the best comprehensively,with an AUC of 0.851(95%CI:0.778-0.924)in the training set and 0.808(95%CI:0.700-0.917)in the external validation set,and provided superior clinical net benefit compared with the single model and the portal venous phase combined model.Conclusion The combined model constructed based on the arterial phase can predict treatment response after TACE in patients with HCC to a certain extent,providing auxiliary reference for clinical treatment decision-making and prognostic assessment.

牛童童;孙新圩;杨盼;穆伟斌

齐齐哈尔医学院 医学技术学院,黑龙江 齐齐哈尔 161000齐齐哈尔医学院 医学技术学院,黑龙江 齐齐哈尔 161000齐齐哈尔医学院 医学技术学院,黑龙江 齐齐哈尔 161000齐齐哈尔医学院 研究生院,黑龙江 齐齐哈尔 161000

肝细胞癌计算机断层扫描治疗反应经动脉化疗栓塞术

hepatocellular carcinomacomputed tomographytreatment responsetransarterial chemoembolization

《分子影像学杂志》 2026 (8)

1044-1050,7

黑龙江省高等教育教学改革研究项目(SJGZB2024219)齐齐哈尔市科技计划联合引导项目(LSFGG-2026116)齐齐哈尔医学科学院自选科研基金项目(QMSI2024Z-20)

10.12122/j.issn.1674-4500.2026.08.09

评论