CT影像组学联合血清参数可有效预测肝细胞癌导管动脉化疗栓塞术后的早期复发  

CT radiomics combined with serum parameters can effectively predict early recurrence after transcatheter arterial chemoembolization for liver cancer

作　　者：孙志刚 雷学迪 孟涛 胡宇康 张宁 杨鹏辉 童钟[3] SUN Zhigang;LEI Xuedi;MENG Tao;HU Yukang;ZHANG Ning;YANG Penghui;TONG Zhong(Anhui Medical University,Hefei 230032,China;BengBu Medical University,Bengbu 233030,China;Department of Hepatobiliary Surgery,Affiliated Hefei First People's Hospital,Hefei 230036,China;Department of Imaging Center,Affiliated Hefei Binhu Hospital,Hefei 230092,China;College of Public Health,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]安徽医科大学,安徽合肥230032 [2]蚌埠医科大学,安徽蚌埠233030 [3]合肥市第一人民医院肝胆外科,安徽合肥230036 [4]合肥市滨湖医院影像科,安徽合肥230092 [5]郑州大学公共卫生学院,河南郑州450001

出　　处：《分子影像学杂志》2025年第3期253-263,共11页Journal of Molecular Imaging

基　　金：安徽省高等学校科学研究项目(AHWJ2023BAa20181);安徽省卫生健康科研项目(2024AH050782)。

摘　　要：目的 本研究旨在开发一种基于增强CT图像特征联合临床指标的联合列线图模型来预测未接受治疗的中期肝细胞癌(HCC)患者经导管动脉化疗栓塞(TACE)的早期复发(ER),并将该模型的性能与影像组学和临床模型进行比较。方法 本研究为回顾性、双中心研究,纳入2020年2月~2024年2月于合肥市第一人民医院就诊的55例在TACE前接受增强CT的HCC患者,通过5折交叉验证随机分为训练组和验证组。收集、评估所有患者的临床资料、CT影像组学数据、病理学资料、TACE术前1周内血清学指标。使用单因素秩和检验和Spearman相关性分析法进行影像组学特征筛选、根据逻辑回归模型系数和特征数值的线性乘积计算Radscore,利用单多因素逻辑回归进行变量显著性分析,得到独立影响因素并建立列线图、绘制ROC曲线及决策曲线以评估模型性能。结果 在训练组和验证组中,联合列线图模型预测TACE术后早期复发的AUC分别为0.787(95%CI:0.52~1.05)和0.847(95%CI:0.54~1.14)。单、多因素回归分析表明凝血酶原时间是与TACE术后早期复发相关的独立血清学影响因素(P<0.05)。在训练组及验证组中,单纯临床模型、单纯影像列线图模型预测TACE早期复发的曲线下面积(AUC)、准确度、敏感度、特异度均低于临床-影像联合列线图模型,通过决策曲线分析表明联合列线图模型具有更大的净收益。结论 本研究所提出的联合列线图模型具有准确预测HCC患者行TACE术后早期复发的潜力。Objective To develop a combined nomogram model based on enhanced CT imaging features and clinical indicators for predicting early recurrence(ER)in untreated intermediate-stage hepatocellular carcinoma(HCC)patients after transcatheter arterial chemoembolization(TACE),and to compare the performance of this model with radiomics and clinical models.Methods In this retrospective,two-center study,55 HCC patients who underwent enhanced CT before TACE at Affiliated Hefei First People's Hospital from February 2020 to February 2024 were randomly divided into training and validation groups using five-fold cross-validation.Clinical data,CT radiomics data,pathological data,and serum markers collected within one week before TACE were collected and evaluated for all patients.Radiomic features were selected using univariate rank sum tests and Spearman correlation analysis.Radscore was calculated based on the linear product of logistic regression model coefficients and feature values.Significant variables were identified using univariate and multivariate logistic regression,and a nomogram was constructed.The model's performance was assessed using ROC curves and decision curve analysis.Results In both the training and validation groups,the combined nomogram model had AUCs of 0.787(95%CI:0.52-1.05)and 0.847(95%CI:0.54-1.14),respectively,for predicting early recurrence after TACE.Univariate and multivariate regression analysis indicated that prothrombin time(P<0.05)was an independent serum marker associated with early recurrence after TACE.In both the training and validation groups,the AUC,accuracy,sensitivity,and specificity of the clinical model and the radiomic nomogram model alone were lower than those of the combined clinical-radiomic nomogram model.Decision curve analysis showed that the combined nomogram model had greater net benefit.Conclusion The proposed combined nomogram model has the potential to accurately predict early recurrence in HCC patients after TACE.

关 键 词：影像组学 列线图 肝细胞癌 经导管动脉化疗栓塞 早期复发 

分 类 号：R73[医药卫生—肿瘤]