CT图像影像组学模型在肺栓塞诊断中的应用价值  

The application value of CT image imagingomics in the diagnosis of pulmonary embolism

作　　者：阮小伟 杨彦兵 吴林桦 唐路松 汪芳[1,2] 杨利莉[1,2] 平学军[2] 任嘉梁 RUAN Xiaowei;YANG Yanbing;WU Linhua;TANG Lusong;WANG fang;YANG Lili;PING Xuejun;REN Jialiang(People's Hospital of Ningxia Hui Autonomous Region,Yinchuan 750002,China;Affiliated People's Hospital of Ningxia Medical University,Yinchuan 750004,China;General Hospital of Ningxia Medical University,Yinchuan 750004,China;Ningxia Medical University,Yinchuan 750004,China;Beijing GE Pharmaceutical(Shanghai)Co.,LTD.,Shanghai 201203,China)

机构地区：[1]宁夏回族自治区人民医院,宁夏银川750004 [2]宁夏医科大学附属自治区人民医院,宁夏银川750004 [3]宁夏医科大学总医院,宁夏银川750000 [4]宁夏医科大学,宁夏银川750000 [5]北京通用电气药业(上海)有限公司,上海201203

出　　处：《宁夏医学杂志》2025年第2期115-118,F0003,共5页Ningxia Medical Journal

摘　　要：目的探讨基于CT图像的影像组学特征预测模型在肺栓塞诊断中应用的可行性及应用价值。方法回顾性收集肺栓塞病例56例行CT肺动脉血管造影(CTPA)的肺栓塞病例,所有入组病例均于肺动脉主干或其分支见到明显的栓子。使用ITK-SNAP软件选取栓塞动脉供血区域最大层面以及对侧正常对应位置,手动勾画感兴趣区(ROI)。对56例肺栓塞病例按照7∶3比例随机分成训练集(40例)和验证集(16例)。使用后PyRadiomics软件进行影像特征提取;经过特征降维后建立logistic回归模型;以交叉验证的方法对逻辑回归模型进行检验。采用ROC曲线下面积去评价独立预测因素的诊断效能。结果根据训练集和验证集的不同,共提取660个影像组学特征,通过特征筛选最终得到3个影像组学特征最具有预测价值,分别为基于直方图特征的峰度、偏度,基于灰度共生矩阵(GLCM)的原变,该模型在训练集、验证集中的AUC值分别为0.779(95%CI:0.68~0.879)、0.742(95%CI:0.57~0.915)。结论基于CT图像的影像组学特征预测模型可为诊断肺栓塞提供有效依据。Objective To explore the feasibility and application value of radiomics feature prediction model based on CT images in the diagnosis of pulmonary embolism.Methods The CT pulmonary angiography(CTPA)images of 56 cases of pulmonary embolism scanned by Gemstone Spectral Imaging(GSI)were retrospectively collected and analyzed.All enrolled cases had obvious emboli in the main pulmonary artery or its branches.ITK-SNAP software was used to select the maximum plane of the embolic artery supply area and the normal corresponding position on the contralateral side,and manually delineate the region of interest(ROI).40 cases were randomly selected as the training group and 16 cases as the validation group.Image feature was extracted by using PyRadiomics software.After feature dimensionality reduction,the logistic regression model was established.The regression model was tested by using cross-validation.The area under the ROC curve was used to evaluate the diagnostic performance of independent predictors.Results According to the differences between the embolization group and the normal group,a total of 660 radiomics features were extracted.Through feature screening,three radiomics features were finally identified as having the most predictive value.kurtosis and skewness based on histogram features,and the AUC values of the model in the training and validation sets were 0.779(95%CI:0.68~0.879)and 0.742(95%CI:0.57~0.915),respectively.Conclusion The radiomics feature prediction model based on CT images can provide effective basis for the diagnosis of pulmonary embolism.

关 键 词：肺栓塞 影像组学 预测模型 肺动脉血管造影 

分 类 号：R563.5[医药卫生—呼吸系统]