基于动脉期CT图像特征预测食管鳞状细胞癌脉管瘤栓状态的初步研究  

Preliminary Study of Arterial Phase CT Imaging Features in Predicting Vascular Tumor Embolus in Esophageal Squamous Cell Cancer

作　　者：李扬[1] 苏海燕 谷霄龙 杨丽[1] 王向明[1] 岳萌[2] 王明博[3] 苏晓华[4] 任嘉良 时高峰[1] LI Yang;SU Haiyan;GU Xiaolong(Department of CT Magnetic Resonance,the Fourth Hospital of Hebei Medical University,Shijiazhuang,Hebei Province 050011,P.R.China)

机构地区：[1]河北医科大学第四医院CT磁共振科,石家庄050011 [2]河北医科大学第四医院病理科,石家庄050011 [3]河北医科大学第四医院胸外科,石家庄050011 [4]河北省人民医院肿瘤内科,石家庄050011 [5]GE中国,北京100176

出　　处：《临床放射学杂志》2022年第9期1699-1706,共8页Journal of Clinical Radiology

摘　　要：目的初步探讨基于动脉期CT图像特征在预测食管鳞状细胞癌脉管瘤栓状态中的价值。方法回顾性分析197例经手术切除治疗的食管鳞状细胞癌患者,所有患者均在术前2周内行动脉期增强CT检查。将所有入组的患者分为脉管瘤栓阳性组与脉管瘤栓阴性组,并按照7∶3的比例将患者随机分为训练集与测试集。使用薄层1.0 mm动脉期CT图像分析肿瘤特征,包括CT强化相关特征和CT形态相关特征分析。CT强化相关特征分析包括肿瘤CT值(CT值_(肿瘤))、正常食管CT值(CT值_(正常))、两者的差值(CT值_(差值))及比值(CT值_(比值))、肿瘤坏死、肿瘤内增粗小血管。CT形态相关特征分析包括肿瘤最大厚度(CT_(厚度))、肿瘤最大长度(CT_(长度))、肿瘤体积(CT_(体积))、肿瘤边缘形态。在训练集中,使用单因素逻辑回归分析筛选出与脉管瘤栓状态相关的CT特征,并纳入多因素逻辑回归分析。使用多因素逻辑回归分析筛选出脉管瘤栓状态的独立预测因素,并分别对各独立预测因素及其联合模型进行受试者工作特征曲线及诊断效能分析。结果单因素逻辑回归分析显示,CT值_(肿瘤)、CT值_(差值)、CT值_(比值)、肿瘤坏死、肿瘤内增粗小血管、CT_(厚度)、CT_(长度)、CT_(体积)、肿瘤边缘形态与脉管瘤栓状态呈正相关。多因素逻辑回归分析显示,CT值_(比值)、CT_(厚度)及肿瘤边缘形态是食管鳞状细胞癌脉管瘤栓状态的独立预测因素。受试者工作特征曲线分析显示,包含3个独立预测因素的联合模型的曲线下面积最高(训练组=0.832,95%CI:0.755~0.909;测试组=0.821,95%CI:0.705~0.937)。结论基于动脉期CT图像特征能够有效预测食管癌鳞状细胞癌脉管瘤栓状态,可以据此区分脉管瘤栓阳性高风险人群。Objective To preliminarily investigate the value of arterial phase CT imaging features in predicting vascular tumor embolus status in esophageal squamous cell cancer(ESCC). Methods This retrospective study included 197 patients with ESCC treated by surgical resection, all of whom underwent arterial phase enhanced CT within 2 weeks prior to surgery.All enrolled patients were divided into vascular tumor embolus positive and negative groups and randomized into training and test sets at a 7∶3 ratio.Tumor characterization, including CT enhancement correlation analysis and CT morphology analysis, was performed using thin layer 1.0 mm arterial phase CT images.The analysis of CT enhancement related features included tumor CT attenuation values(CTtumor),normal esophageal CT attenuation values(CTnormal),the difference(CTdifference) and ratio(CTratio) between them, tumor necrosis, and enlarged small vessels within tumor(ESVWT).CT morphology related features including maximum tumor thickness(CTthick),maximum tumor length(CTlength),tumor volume(CTvolume),and tumor margin morphology were analyzed.In the training set, CT features associated with vascular tumor embolus status were filtered out using univariate logistic regression analysis and then included in the multivariate analysis.Independent predictors of LVI status were screened using multivariate logistic regression analysis, and ROC curves and diagnostic efficacy analyses were performed for each independent predictor and their combination, respectively. Results Univariate logistic regression analysis showed that CTtumor, CTdifference,(CTratio),tumor necrosis, ESVWT,CTthick, CTlength, CT volume and tumor margin morphology were associated with LVI status.Multivariate logistic regression analysis showed that CTratio, CTthick and tumor margin morphology were independent predictors of LVI status in esophageal cancer.ROC curve analysis showed the highest AUC values for the combined model containing three independent predictive features(training set = 0.832,95% CI:0.755-0.909

关 键 词：食管鳞状细胞癌 脉管瘤栓 体层摄影术 X线计算机 增强 动脉期 

分 类 号：R735.1[医药卫生—肿瘤] R730.44[医药卫生—临床医学]