CT影像组学特征预测胰管结石体外冲击波碎石术疗效的价值

Value of CT imaging radiomics in predicting the clinical efficacy of extracorporeal shock wave lithotripsy for pancreatic duct stones

目的探讨CT影像组学特征在预测胰管结石体外冲击波碎石(ESWL)治疗效果中的价值。方法回顾性分析2016年7月至2023年1月间西湖大学附属杭州市第一人民医院消化内科167例行ESWL治疗的胰管结石患者的临床资料。根据首次ESWL治疗后最大残留结石的大小将患者分为完全碎石组(结石直径≤3 mm,94例)和不完全碎石组(结石直径>3 mm,73例)。使用ITK-SNAP软件对胰管结石图像进行勾画,利用联影公司开发的人工智能工具包提取影像组学特征,将胰管结石数据集按照8∶2比例随机分配到训练集(118例)和测试集(29例)中,采用绝对值最大归一化处理,再通过最小绝对收缩和选择算子(Lasso)进行降维和选择,计算CT影像组学分值,逻辑回归分类器构建胰管结石ESWL治疗效果预测模型。绘制受试者工作特征曲线(ROC),计算曲线下面积(AUC)及灵敏度、特异度、准确率,以评估预测模型的性能。采用决策曲线分析评估CT组学分值诊断胰管结石ESWL疗效的临床应用价值。结果共提取2287个影像组学特征,经Lasso回归降维最终筛选11个最佳影像组学特征建立胰管结石ESWL治疗效果的预测模型,其训练集和测试集的AUC值分别为0.89、0.87,灵敏度、特异度、准确率分别为82%、79%,82%、82%,82%、80%。独立验证集中AUC值为0.90,灵敏度、特异度、准确率分别为78%、90%、85%。决策曲线分析结果显示,当用CT影像组学分值诊断胰管结石ESWL疗效的概率>0.05时,使用CT影像组学分值诊断胰管结石ESWL疗效比不使用更能使患者在临床中获益。结论应用CT影像组学特征模型可以预测胰管结石ESWL的治疗效果。

Objective To investigate the value of CT imaging radiomics in predicting the therapeutic effect of extracorporeal shock wave lithotripsy(ESWL)for pancreatic duct stones.Methods The clinical data of 167 patients with pancreatic duct stones treated with ESWL in the Department of Gastroenterology,the First People's Hospital of Hangzhou,Westlake University from July 2016 to January 2023 were retrospectively analyzed.Patients were divided into complete lithotripsy group(stone diameter≤3 mm,n=94)and incomplete lithotripsy group(stone diameter>3 mm,n=73),according to the size of the largest residual stone after the first ESWL treatment.ITK SNAP software was used to delineate the images of pancreatic duct stones,and the artificial intelligence tool kit developed by United Shadow Company was used to extract the image radiomics characteristics.The pancreatic duct stone data set was randomly assigned into the training set(n=118)and the test set(n=29)in the ratio of 8∶2,and the absolute maximum normalization treatment was used,followed by peacekeeping selection through the minimum absolute contraction and selection operator(Lasso)to calculate the CT image radiomics score,and the logistic regression classifier was used to construct the ESWL treatment effect prediction model of pancreatic duct stones.Receiver operating characteristic curves(ROC)were plotted,and the area under the curve(AUC)and sensitivity,specificity,and accuracy were calculated to assess the performance of the prediction model.Decision curve analysis was used to evaluate the clinical value of CT radiomics score in the diagnosis of ESWL for pancreatic duct stones.Results A total of 2287 imaging radiomics characteristics were extracted,and 11 optimal imaging radiomics characteristics were finally screened by Lasso regression dimensionality reduction to establish a prediction model for ESWL treatment effect of pancreatic duct stones.The AUC values of the training set and the test set were 0.89 and 0.87,respectively,and the sensitivity,specificity,and accuracy were 82%and 79%,82%and 82%,82%and 80%,respectively.The AUC value in the independent validation set was 0.90,and the sensitivity,specificity,and accuracy were 78%,90%,and 85%,respectively.The results of decision curve analysis showed that when the probability of ESWL efficacy in the diagnosis of pancreatic duct stones with CT image radiomics score was>0.05,the use of CT image radiomics score in the diagnosis of ESWL efficacy in pancreatic duct stones was more beneficial to patients in clinical practice than not.Conclusions The treatment effect of ESWL for pancreatic duct stones can be predicted by CT imaging radiomics model.