基于增强CT门静脉期图像的影像组学模型对进展期胃腺癌浆膜浸润的预测价值  

The Predictive Value of Imageomics Model Based on Enhanced CT Portal Venous Phase Images for Serosal Infiltration in Advanced Gastric Adenocarcinoma

作　　者：田洪滢 徐军[1] Tian Hongying;Xu Jun(Northern Jiangsu People's Hospital Affiliated to Yangzhou University,Yangzhou Jiangsu 225001,China;Yangzhou Second People's Hospital,Yangzhou Jiangsu 225007,China)

机构地区：[1]扬州大学附属苏北人民医院,江苏扬州225001 [2]扬州市第二人民医院,江苏扬州225007

出　　处：《医疗装备》2023年第6期17-20,共4页Medical Equipment

摘　　要：目的 利用uAI科研平台建立进展期胃腺癌增强CT的影像组学特征模型,探究其预测浆膜浸润的可行性。方法 回顾性分析2019年1月至2021年12月扬州大学附属苏北人民医院收治的170例进展期胃腺癌患者的临床资料,其中浆膜浸润组90例,非浆膜浸润组80例。所有患者均行增强CT扫描,在增强CT门静脉期图像上勾画感兴趣区(ROI)并提取影像组学特征,采用五折交叉(种子数为20)法将170例患者的影像数据分为训练组(136例)和测试组(34例),采用LASSO算法实现高维数据的降维,并用逻辑回归分类器构建预测进展期胃腺癌浆膜浸润的影像组学模型,最后绘制受试者工作特征曲线(ROC曲线),用曲线下面积(AUC)、精确率、灵敏度、特异度及准确率评估模型的预测效能。结果 进展期胃腺癌门静脉期影像组学模型训练组和测试组AUC值分别为0.868、0.823,精确率、灵敏度、特异度及准确率分别为0.808、0.794、0.788、0.791和0.752、0.767、0.725、0.747。结论 基于增强CT门静脉期图像的影像组学模型对进展期胃腺癌浆膜浸润具有较好的预测价值。Objective The imaging feature model of enhanced CT of advanced gastric adenocarcinoma was established by using the uAI scientific research portal,and the feasibility of its prediction of serosal infiltration was explored.Methods With the retrospective analysis of the clinical data of 170 patients with advanced gastric adenocarcinoma admitted to the Subei People's Hospital affiliated to Yangzhou University from January 2019 to December 2021,including 90 patients in serosal infiltration group and 80 patients in nonserosal infiltration group,all patients underwent enhanced CT scanning.On the enhanced CT portal phase image,the region of interest(ROI)was delineated,and the imaging features were extracted.The image data of 170 patients were divided into training group(136 cases)and test group(34 cases),using the five-fold cross validation(seed number of 20).The dimensionality reduction of high-dimensional data was realized by using the LASSO algorithm,and the imageomics model for predicting the serosal infiltration of advanced gastric adenocarcinoma was constructed by using the logistic regression classifier.Finally,the receiver operating characteristic curve(ROC curve)was drawn,and the prediction efficiency of the model was evaluated by the area under the curve(AUC),accuracy,sensitivity,specificity and accuracy.Results The AUC values training group and test group of advanced gastric adenocarcinoma in portal venousphase imaging model were 0.868 and 0.823, respectively;The accuracy, sensitivity, specificityand accuracy were 0.808, 0.794, 0.788, 0.791 and 0.752, 0.767, 0.725 and 0.747, respectively.Conclusion The imageomics model based on enhanced CT portal venous phase image hasgood predictive value for serosal infiltration of advanced gastric adenocarcinoma.

关 键 词：胃腺癌 浆膜浸润 CT 影像组学 

分 类 号：R735.2[医药卫生—肿瘤]