基于CT影像组学对局灶性自身免 疫性胰腺炎与胰腺导管腺 癌鉴别诊断的价值

The value of CT imaging omics in the differentiation of focal autoimmune pancreatitis and pancreatic cancer

目的:探讨基于CT影像组学对局灶性自身免疫性胰腺炎(focal-autoimmune pancreatitis,f-AIP)与胰腺导管腺癌(pancreatic ductal adenocarcinoma,PDAC)鉴别诊断的价值。方法:回顾性收集f-AIP患者73例,PDAC患者74例的CT影像资料。首先,由两位放射科医生对所有f-AIP与PDAC患者的CT征象进行分析,确定病变的类型,得出人工诊断的准确性。其次,将显示病灶的所有层面的三期薄层图像以DICOM格式上传至达尔文医准科研平台,在病灶的中心层面及其上下层面勾画感兴趣区(ROI),提取组学特征,运用方差选择法、最小绝对收缩算法(LASSO)对组学特征进行筛选和降维,建立支持向量机(SVM)组学模型,并建立CT征象联合组学特征的联合模型,最后使用受试者工作特性曲线(ROC曲线)对各模型进行评价。结果:放射科医生诊断的准确性为0.83,在平扫、动脉期及静脉期中筛选出5、11、6个组学特征建立模型,联合CT征象组建联合模型,分别筛选出14、8、8个特征。在组学分析中,基于平扫、动脉期及静脉期所建组学模型,静脉期组学模型敏感度0.96、特异度0.90、准确度0.93及AUC值0.97均高于其他两期;在联合模型中,各期联合模型均较同期单纯组学模型的AUC值、敏感度及准确度高,仅平扫单纯模型的特异度高于联合组学模型。结论:在f-AIP与PDAC的鉴别中,人工诊断具有一定的诊断能力;构建多期相SVM单纯组学模型,动脉期及静脉期组学模型诊断效能均高于人工诊断,且静脉期诊断价值最高;采用CT征象联合组学特征建立联合模型,诊断效能明显优于单一模型,进一步提高了对f-AIP与PDAC的鉴别能力。

Objective:To explore the value of multiphase CT imaging omics in the differentiation of autoimmune pancreatitis(AIP)and pancreatic ductal adenocarcinoma(PDAC).Methods:73 cases of f-AIP patients and 74 cases of PDAC patients,with CT images,were collected respectively.Firstly,the CT signs of f-AIP and PDAC were evaluated and identified through imaging data by two radiologists.Secondly,all levels of three-phase lesions were uploaded to the Darwin Medical Research Platform,then the ROI on the central level of the lesion and its upper and lower level were outlined and extracted radiomics features.By using the variance selection method and least absolute shrinkage and selection operator(LASSO)to screen out the radiomics characteristics,three phases of Support Vector Machine(SVM)were established.And establish a joint model of CT signs and omics features.Finaly,through evaluated each model using ROC,the best phase diagnosis model was obtained and compared with artificial diagnosis results.Results:The diagnostic accuracy of radiologists was 0.83.In the omics analysis,5,11 and 6 imaging omics features were screened out of in unenhanced phase,arterial phase and venous phase respectiveiy,and in the joint model,14,8,8 features were screened out respectively.Among the omics models,the diagnostic accuracy,sensitivity,specificity,efficiency of the venous phase are the highest(0.96,0.9,0.93 and 0.97 respectively).In joint model,each phase model has higher AUC value,sensitivity and accuracy than the omics model of the same period.Conclusion:In the identification of f-AIP and PDAC,the CT signs has a certain diagnostic ability.To construct a multi-phase SVM omics model,the diagnostic efficiency of the arterial and venous omics models is higher than the radiologists,and the diagnostic value of the venous phase is the highest.The diagnostic efficiency of joint model is significantly better than that of a single model.