光谱CT细胞外容积临床-影像组学模型预测结直肠癌血管淋巴管及神经侵犯  

Spectral CT Extracellular Volume Clinical-radiomics Models for Prediction of Vascular Lymphatic and Nerve Invasion in Colorectal Cancer

作　　者：郑文霞 王莉莉[2] 陈杏彪 郑小霞 崔雅琼 黄刚[2] ZHENG Wenxia;WANG Lili;CHEN Xingbiao;ZHENG Xiaoxia;CUI Yaqiong;HUANG Gang(The First Clinical College of Medicine,Gansu University of Chinese Medicine;Department of Radiology,Gansu Provincial Hospital,Gansu,China,Lanzhou 730000,China;Philips Medical Clinical Research Department,Shanghai 200070,China)

机构地区：[1]甘肃中医药大学第一临床医学院 [2]甘肃省人民医院放射科,兰州730000 [3]飞利浦医疗临床科研部

出　　处：《中国医学计算机成像杂志》2023年第5期551-559,共9页Chinese Computed Medical Imaging

基　　金：甘肃省青年基金计划项目(20JR5RA143)。

摘　　要：目的:探讨基于光谱CT细胞外容积参数图影像组学特征联合临床特征来预测结直肠癌的血管淋巴管和神经侵犯,以期获得一种术前无创评估的方法。方法:回顾性收集155例经术后病理证实的结直肠癌患者。采用ITK-SNAP软件进行半自动分割勾画肿瘤三维感兴趣区(VOI)。采用FeAture Explorer (FAE)软件提取特征,并进行特征筛选、组学模型构建。在临床指标中选择有统计学意义的指标建立临床模型,由临床指标及组学特征进行logistic回归分析建立联合模型。利用受试者工作特征曲线(ROC)及曲线下面积(AUC)来评价3种模型预测结直肠癌血管淋巴管及神经侵犯的价值。结果:血管侵犯组中临床模型训练集与测试集的AUC分别为0.613、0.521;组学模型训练集与测试集的AUC分别为0.797、0.852;联合模型训练集与测试集的AUC分别为0.809、0.862。淋巴管侵犯组中临床模型训练集与测试集的AUC分别为0.636、0.601;组学模型训练集与测试集的AUC分别为0.768、0.806;联合模型训练集与测试集的AUC分别为0.785、0.823。神经侵犯组中临床模型训练集与测试集的AUC分别为0.712、0.771;组学模型训练集与测试集的AUC分别为0.800、0.819;联合模型训练集与测试集的AUC分别为0.835、0.892。结论:基于影像细胞外容积的影像组学特征联合临床特征构建模型可有效预测结直肠癌的血管淋巴管及神经侵犯,具有术前无创诊断价值。Purpose:To investigate the prediction of vascular lymphatic and nerve invasion in colorectal cancer based on spectral CT extracellular volume parameter map radiomics features combined with clinical features,with the aim of obtaining a method for preoperative noninvasive assessment.Methods:A retrospective collection of 155 patients with postoperative pathologically confirmed colorectal cancer was made.Semi-automatic segmentation was performed using ITK-SNAP software to outline the tumor 3D volume of interest(VOI).FeAture Explorer(FAE)software was used to extract features,and feature screening and radiomics model construction were performed.The statistically significant clinical indicators were selected among the clinical indicators to build the clinical models,and logistic regression analysis was performed from the clinical indicators and radiomics features to build the combined models.The receiver operating characteristic(ROC)curve and area under the curve(AUC)were used to evaluate the value of the three models in predicting vascular lymphatic and neural invasion in colorectal cancer.Results:In the vascular invasion group,the AUC of the training set and the test set of the clinical model were 0.613 and 0.521 respectively.The AUC of the training set and the test set of the radiomics model were 0.797 and 0.852,respectively.The AUC of the combined model training set and the test set were 0.809 and 0.862,respectively.In the lymphatic vessel invasion group,the AUC of training set and test set of the clinical model were 0.636 and 0.601,respectively.The AUC of the training set and the test set of the radiomics model were 0.768 and 0.806,respectively.The AUC of the combined model training set and the test set were 0.785 and 0.823,respectively.In the neural group,the AUC of the training set and the test set of the clinical model were 0.712 and 0.771,respectively.The AUC of the training set and the test set of the radiomics model were 0.800 and 0.819,respectively.The AUC of the combined model training set and the test set

关 键 词：能谱CT 结直肠癌 细胞外容积 影像组学 血管淋巴管及神经侵犯 

分 类 号：R445.3[医药卫生—影像医学与核医学]