CT纹理分析列线图诊断结肠癌肠外浸润的价值  

Nomogram Based on CT Texture Analysis in Diagnosing Pericolic Infiltration of Colorectal Carcinoma

作　　者：王晗[1] 陈雨清 郭盛仁[3] 陈光强[3] WANG Han;CHEN Yuqing;GUO Shengren;CHEN Guangqiang(Department of Diagnostic Imaging,the Second Affiliated Hospital of Soochow University,Suzhou 215008,China;不详)

机构地区：[1]南京大学医学院附属鼓楼医院超声诊断科,江苏南京210009 [2]徐州医科大学附属医院超声科,江苏徐州221000 [3]苏州大学附属第二医院影像诊断科,江苏苏州215008

出　　处：《中国医学影像学杂志》2021年第7期709-713,共5页Chinese Journal of Medical Imaging

摘　　要：目的探讨CT纹理分析列线图(Nomogram模型)鉴别结肠癌肠外浸润的价值。资料与方法纳入术前1周行腹盆腔CT扫描示肠周具有浸润征象、手术病理证实为结肠癌的100例患者,其中癌性浸润50例,炎性浸润50例。比较两组肠周脂肪直方图及灰度共生矩阵纹理特征的差异,绘制受试者工作特征(ROC)曲线求截断值,纳入多因素Logistic回归模型,绘制列线图(Nomogram)、ROC曲线图及校准曲线图。结果癌性侵犯组与炎性浸润组静脉期百分位数(P90)、静脉期熵、延迟期变异数差比较,差异均有统计学意义(P<0.001);各特征的ROC曲线下面积(AUC)均>0.70,以静脉期熵鉴别效能最佳(AUC为0.89)。上述纹理特征纳入多因素Logistic回归模型绘制列线图将模型可视化,模型诊断准确度为84%,AUC为0.91,校准曲线图显示预测癌性浸润和实际癌性浸润一致性较高。结论基于CT纹理分析的多因素Logistic回归建立Nomogram预测模型在鉴别结肠癌肠周脂肪浸润征象方面有较高的效能。Purpose To explore the value of Nomogram based on CT texture analysis in differentiating periodic infiltration of patient with colorectal carcinoma.Materials and Methods A total of 100 patients who were diagnosed as colorectal carcinoma through postoperative pathological examination and received abdominal and pelvic three-phase contrast CT scans one week before operation were retrospectively analyzed.All patients were divided into two groups,including 50 patients with periodic cancerous infiltration and 50 patients with inflammatory infiltration.The differences of histogram and gray level co-occurrence matrix(GLCM)texture features of periodic adipose tissue were compared between these two groups.The receiver operating characteristic(ROC)curve was drawn to calculate the cut-off value.A multivariate logistic regression was further established,calculating the Nomogram,ROC curve and calibration graph,etc.Results There were statistically significant differences in the percentile of P90(venous phase),entropy(venous phase)and the variation(delay phase)between the group with cancerous invasion and the group with inflammatory invasion(P<0.001).The area under the curve(AUC)of each feature was all>0.70,showing the optimal identification efficiency in entropy of venous phase(the AUC was 0.89).A multivariate logistic regression model including these three texture features were performed to visualize the model via Nomogram.The diagnostic accuracy rate was 84%,and the AUC of ROC was 0.91,respectively.The calibration curve showed a high consistency in cancerous infiltration between predicted group and actual group.Conclusion The quantitative information provided by multivariate Logistic regression model via Nomogram based on CT texture analysis could be used for differentiating periodic cancerous infiltration in patient with colorectal carcinoma.

关 键 词：结肠肿瘤 体层摄影术 螺旋计算机 T分期 纹理分析 列线图 病理学 外科 诊断 鉴别 

分 类 号：R730.42[医药卫生—肿瘤] R735.3[医药卫生—临床医学]