CT影像组学联合炎症指标构建逻辑回归模型预测食管鳞癌新辅助化疗疗效  

Logic regression model of CT-based radiomics combined with hematological inflammatory features forpredicting the efficacy of neoadjuvant chemotherapy in patients with esophageal squamous carcinoma

作　　者：宫悦 胡逸凡 夏茜[1] 丁娟娟[1] 顾红梅[2,3] GONG Yue;HU Yi-fan;XIA Qian(Department of Radiology,the Dongtai People's Hospital,Jiangsu 224200,China)

机构地区：[1]东台市人民医院影像科,江苏东台224200 [2]南通大学附属医院影像科,江苏南通226000 [3]南通大学医学院,江苏南通226000

出　　处：《放射学实践》2022年第12期1474-1479,共6页Radiologic Practice

摘　　要：目的:探究基于CT影像组学联合血液学炎症指标构建逻辑回归模型预测食管鳞癌新辅助化疗(NAC)疗效的可行性。方法:回顾性分析两家医院经病理证实的54例食管鳞癌患者在术前规范化NAC前、后两次胸部CT增强图像及NAC前一周内的血液学炎症指标检测结果。测量治疗前、后病灶的最长径,计算其变化率,并根据实体肿瘤疗效评价标准(RECIST 1.1),将患者分为NAC有效组(30例)及无效组(24例)。采用独立样本t检验或Mann-Whitney U检验筛选血液学炎症指标中与疗效相关的因素。在患者治疗前静脉期图像上沿肿瘤边界逐层手工勾画ROI,最终生成三维感兴趣区(VOI)并提取其影像组学特征,使用最小冗余最大相关及Boruta工具包进行特征筛选并构建影像组学标签。分别建立影像组学特征、血液学炎症指标、影像组学标签联合血液学炎症指标的逻辑回归模型,采用混淆矩阵和ROC曲线分析模型对NAC疗效的预测效能,采用DCA曲线评估其临床实用价值。结果:外周血淋巴细胞计数及淋巴细胞数与单核细胞数的比值被纳入炎症指标模型。于治疗前静脉期图像上共提取了1168个组学特征,经降维后共筛选出5个影像组学特征(wavelet-HLL_gldm_DependenceEntropy、wavelet-HHL_gldm_LargeDependenceLowGrayLevelEmphasis、wavelet-HHH_glrlm_HighGrayLevelRunEmphasis、wavelet-HHH_glrlm_LowGrayLevelRunEmphasis和wavelet-HLL_glszm_ZoneEntropy)用于构建影像组学标签。基于影像组学、血液学炎症指标以及联合模型预测NAC疗效的的AUC分别为0.77、0.72和0.80。结论:基于新辅助化疗前的增强CT影像组学及血液学炎症指标特征构建的预测模型可较好的预测食管鳞癌患者新辅助化疗疗效,以联合模型的效能最优,可为临床制订个性化治疗方案提供参考。Objective:To investigate the feasibility of constructing a logic regression model based on radiomics signature extracted from CT images combined with clinical hematological inflammatory features to predict the efficacy of neoadjuvant chemotherapy in patients with esophageal squamous carcinoma.Methods:A total of 54 patients with pathologically confirmed esophageal squamous carcinoma who received standardized neoadjuvant chemotherapy before surgery in two hospitals with pre-and post-treatment chest CT-enhanced images and hematological inflammatory features measured before neoadjuvant chemotherapy within one week were retrospectively analyzed.The longest length of each tumor before and after treatment was measured,and then the change rate was calculated.According to the efficacy evaluation criteria of solid tumors(RECIST 1.1),the patients were divided into NAC effective group(30 cases)and ineffective group(24 cases).Factors associated with efficacy in hematological inflammatory features were screened by independent sample t-test or Mann-Whitney U-test.ROIs were manually delineated slice by slice along the tumor boundary on venous phase images of pre-treatment patient,which finally the VOI was generated and its radiomics features were extracted,and the Boruta toolkit for feature was used to extract and construct radiomics score.Logic regression mo-dels of radiomics features,clinical hematological inflammatory indexes and clinical-radiomics indexes were established respectively.The predictive performance for NAC of the models was evaluated by confusion matrix and ROC curve,and its practical value in clinic was evaluated by DCA curve.Results:The ratio of peripheral blood lymphocyte count,lymphocyte count to monocyte count was included in the clinical model.A total of 1,168 radiomics features were extracted from the pre-treatment venous phase images.After dimension reduction,five radiomics features including wavelet-HHL_gldm_LargeDependenceLowGrayLevelEmphasis,wavelet-HLL_gldm_DependenceEntropy,wavelet-HHH_glrlm_HighGra

关 键 词：食管肿瘤 新辅助化疗 疗效 影像组学 血液学炎症指标 

分 类 号：R814.42[医药卫生—影像医学与核医学] R735.1[医药卫生—放射医学]