基于CT影像组学胰十二指肠切除术后胰瘘风险预测模型及其价值  

The value of risk model based on CT radiomics in predicting postoperative pancreatic fistula following pancreaticoduodenectomy

作　　者：陈龙江 李微波 汤志刚[1] 李祥攀[2] CHEN Longjiang;LI Weibo;TANG Zhigang;LI Xiangpan(Department of Pancreatic Surgery,Renmin Hospital of Wuhan University,Wuhan 430060,China;Centre of Oncology,Renmin Hospital of Wuhan University,Wuhan 430060,China)

机构地区：[1]武汉大学人民医院胰腺外科,湖北武汉430060 [2]武汉大学人民医院肿瘤中心,湖北武汉430060

出　　处：《肝胆胰外科杂志》2022年第11期667-673,共7页Journal of Hepatopancreatobiliary Surgery

摘　　要：目的探讨基于CT影像组学模型术前预测胰十二指肠切除术后胰瘘(POPF)的应用价值。方法回顾性分析106例接受胰十二指肠切除术患者的临床及腹部CT资料,其中POPF(+)组36例,POPF(-)组70例。采用ITAK-SNAP软件勾画CT图像感兴趣区域(ROI),Python程序的radiomics包进行影像组学特征提取,使用最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归进一步筛选特征、建立影像组学评分(Rad-score),构建影像组学预测模型。然后将临床特征、Rad-score纳入多因素Logistic回归分析,筛选出POPF发生的独立危险因素,构建临床预测模型以及联合影像学组学特征的混合模型。最后采用受试者工作特征曲线(ROC)评估不同模型的预测效能。结果共筛选出7个非零影像组学特征并建立了Rad-score。BMI、胰管扩张及Rad-score均是发生POPF的独立危险因素。影像组学预测模型、临床特征预测模型及混合预测模型预测POPF的曲线下面积(AUC)为分别为0.72、0.69、0.80;Delong检验表明临床特征预测模型与混合预测模型间的差异具有统计学意义。结论基于CT影像组学模型在术前辅助预测胰十二指肠切除术POPF方面具有一定的价值,联合临床指标能够提高模型的预测效能。objective To evaluate the application value of CT-based radiomics model for preoperative prediction of postoperative pancreatic fistula(POPF)in patients undergoing pancreaticoduodenectomy(PD).Methods Clinical and abdominal CT data of 106 patients who underwent PD were retrospectively analyzed.POPF occured in 36 patients.ITAK-SNAP software was used to delineate the region of interest(ROI)of CT images,and radiomics package of Python program was used to extract the radiomics features.The least absolute shrinkage and selection operator(LASSO)regression was used to further select features,construct radiomics score(Rad-score)and radiomics prediction model.Then the clinical features and Rad-score were subjected into multivariate Logistic regression analysis to identify the independent risk factors for POPF.The clinical prediction model and the mixed model combined with Rad-score were build.Finally the receiver operating characteristic(ROC)curves were used to evaluate the predictive effectiveness of different models.Results A total of 7 non-zero radiomics features were screened out and the Rad-score were established.BMI,pancreatic duct dilatation and Rad-score were independent risk factors for POPF.The area under curve(AUC)for predicting POPF in radiomics,clinical and mixed model were 0.72,0.69 and 0.80,respectively.Delong test showed that the difference between the clinical and mixed model was statistically significant.Conclusion The CT-based radiomics model is useful in the preoperative prediction of POPF in patients who underwent PD.The prediction efficiency of the model may be improved by combining with clinical indicators.

关 键 词：影像组学 胰瘘 预测模型 计算机断层扫描(CT) 

分 类 号：R445[医药卫生—影像医学与核医学] R735.9[医药卫生—诊断学]