基于CT影像组学特征术前预测肝细胞癌微血管侵犯的价值  

VALUE OF CT-BASED RADIOMIC FEATURES IN PREOPERATIVE PREDICTION OF MICROVASCULAR INVASION IN HEPATOCELLULAR CARCINOMA

作　　者：何颖 周显军[1] 陈鑫[1] 朱呈瞻 董蒨[1] HE Ying;ZHOU Xianjun;CHEN Xin;ZHU Chengzhan;DONG Qian(Department of Pe-diatric Surgery, The Affiliated Hospital of Qingdao University, Qingdao 266003, China)

机构地区：[1]青岛大学附属医院小儿外科,山东青岛266003 [2]青岛大学附属医院肝胆胰外科,山东青岛266003

出　　处：《精准医学杂志》2022年第1期11-16,21,共7页Journal of Precision Medicine

基　　金：山东省高等学校青创科技支持计划项目(2020-KJL005)。

摘　　要：目的探索基于CT影像组学特征术前预测肝细胞癌微血管侵犯的价值。方法选取我院2013年1月—2018年12月经病理学检查证实的肝细胞癌患者271例,按照3∶1的比例随机分为训练组(203例)和验证组(68例)。从患者的CT图像中提取影像组学特征并用最小冗余最大相关性算法和最小绝对收缩和选择算子算法选择特征,计算影像组学特征评分并构建影像学模型。基于临床危险因素构建临床模型以及构建基于影像组学特征评分和临床危险因素的联合模型,并绘制联合模型的列线图。通过ROC曲线评价各模型预测肝癌微血管是否侵犯的效能,并通过Delong检验比较联合模型与临床模型和影像学模型预测效能是否具有统计学差异。采用校正曲线评估联合模型的拟合度,最后再通过决策曲线评价联合模型预测肝癌微血管侵犯效能的净获益。结果影像组学特征评分是预测肝癌是否发生微血管侵犯的独立因素。联合模型列线图对肝细胞癌显示出良好的预测效能,训练组AUC为0.80(95%CI=0.75~0.85),验证组AUC为0.75(95%CI=0.65~0.85)。联合模型预测肝癌微血管侵犯的效能明显高于影像学模型、临床模型(P<0.05),决策曲线表明联合模型预测肝癌微血管侵犯的效能较临床模型具有较高的净获益。结论基于肝细胞癌患者CT影像组学特征和临床危险因素构建的预测模型,对术前预测肝细胞癌微血管侵犯有一定效能。Objective To investigate the value of CT-based radiomic features in the preoperative prediction of microvascular invasion in hepatocellular carcinoma(HCC).Methods A total of 271 patients with pathologically confirmed HCC who were treated in our hospital from January 2013 to December 2018 were enrolled and randomly divided into training group with 203 patients and validation group with 68 patients at a ratio of 3∶1.Radiomic features were extracted from the CT images of the patients and were selected by the minimum redundancy-maximum relevance algorithm and the least absolute shrinkage and selection operator(LASSO)algorithm.The scores of radiomic features were calculated and then a radiomics model was established.A clinical model was established based on clinical risk factors,and a combined model was established based on the scores of radiomic features and clinical risk factors;a nomogram was plotted for the combined model.The receiver operating characteristic(ROC)curve was used to evaluate the performance of each model in predicting HCC microvascular invasion,and the Delong test was used to compare the predictive performance of the combined model versus the clinical model and the radiomics model.The calibration curve was used to evaluate the degree of fitting of the combined model,and the decision curve was used to evaluate the net benefit of the combined model in predicting HCC microvascular invasion.Results Radiomic score was an independent factor in predicting HCC microvascular invasion.The nomogram of the combined model showed that the model had good performance in predicting HCC,with an AUC of 0.80(95%CI=0.75-0.85)in the training group and 0.75(95%CI=0.65-0.85)in the validation group.The combined model had a significantly higher performance than the radiomics model and the clinical model in predicting HCC microvascular invasion(P<0.05),and the decision curve showed that the combined model had a higher net benefit than the clinical model in predicting HCC microvascular invasion.Conclusion The predictive mode

关 键 词：癌 肝细胞 微血管 肿瘤浸润 体层摄影术 X线计算机 影像组学 列线图 图像处理 计算机辅助 预测模型 

分 类 号：R735.7[医药卫生—肿瘤] R319.1[医药卫生—临床医学]