摘要
目的探讨基于CT平扫的影像组学标签鉴别机化性肺炎与肺腺癌的价值。方法选取池州市人民医院经病理确诊的70例肺腺癌与25例机化性肺炎患者影像资料。在肺窗沿病变所有层面勾画ROI并融合成三维容积感兴趣区(VOI),并提取1050个影像组学特征,包括形状特征、一阶特征、纹理特征,后两者经LoG、小波与LBP滤过处理。使用mRMR、LASSO回归筛选影像组学特征,建立逻辑回归模型。ROC曲线用于评价模型鉴别机化性肺炎及肺腺癌的效能。结果经mRMR、LASSO回归筛选后获得11个最具有鉴别诊断价值的影像组学特征建立影像组学标签。影像组学标签训练组AUC为0.92(95%CI:0.85-0.99),灵敏度、特异度分别为95.82%、73.47%;验证集AUC为0.81(95%CI:0.63-0.99),灵敏度、特异度分别为85.71%、76.19%。结论基于CT平扫的影像组学标签对鉴别机化性肺炎与肺腺癌有一定的支持作用。
Objective To explore the value of CT radiomics signatures indifferentiating organizing pneumonia and lung adenocarcinoma.Methods The image data of 70 patients with lung adenocarcinoma and 24 patients with organizing pneumonia diagnosed pathologically in Chizhou city people’s Hospital were collected.We manually outlined and merged all layers of the lesion along the lung window into a three-dimensional volume of interest(VOI),and extracted 1050 imageomics features,including shape feature,first-order statistical feature and texture feature,the latter two were filtered by laplacian of gaussian,wavelet and local binary pattern.The best feature subset was screened by mRMR,LASSO regression and 10-fold cross verification.The radiomics signature was constructed by the remained fetures and the logical regression was constructed.The ROC curve was used to evaluate the efficiency of the model in distinguishing organizing pneumonia and lung adenocarcinoma.Results After screened by mRMR and LASSO regression,11 radiomics features with the most diagnostic value were obtained.The AUC of the radiomics label training group AUC was 0.92(95%CI 0.85-0.99),the sensitivity and specificity were 95.82%,73.47%respectively;the validation group AUC was 0.81(95%CI 0.63-0.99),the sensitivity and specificity were 85.71%and 76.19%respectively.Conclusion The CT radiomics label based has a certain benefit in distinguishing organizing pneumonia from lung adenocarcinoma.
作者
徐杨飞
俞咏梅
吴琦
刘啸峰
XU Yangfei;YU Yongmei;WU Qi;LIU Xiaofeng(Department of Radiology, Chizhou City People’s Hospital, Chizhou 247100, P.R.China;Department of Radiology, The First Affiliated Hospital of Wannan Medical College, Wuhu 241001, P.R.China)
出处
《医学影像学杂志》
2021年第10期1673-1676,共4页
Journal of Medical Imaging
