Objective To develop prognostic radiomic model of lung adenocarcinoma and to compare the prediction per-formance of different factors and to further investigate the correlation between radiomic features and tumor microenviron-ment.Methods This study was a reanalysis of dataset NSCLC-Radiogenomics in TCIA database.The original study in-cluded 211 non-small cell lung cancer(NSCLC)patients,and 114 lung adenocarcinoma patients were included in model development after inclusion and exclusion.After image segmentation of the patients'preoperative CT,Pyradiomics package was used to extract radiomic features.Combined with disease-free survival data,univariate COX regression-LASSO regres-sion-multivariate COX regression,the successive calculation step,was applied to construct radiomic model predicting postop-erative prognosis of lung adenocarcinoma patients.Patients were divided into high and low risk group according to median risk score,and survival analysis was conducted between the two group.Subsequently,multiple clinical factors were analyzed with model features in multivariate COX regression analysis to obtain the clinical-radiomics model.And prognostic prediction performance was compared among the radiomic model,the mixed model,T stage and G stage by the area under the curve(AUC)of receiver operating characteristic(ROC)curves.Ultimately,ESTIMATE scoring and CIBERSORT analysis results were obtained based on RNA-seq data of tumor samples and furtherly tumor cell content was compared between high and low risk group and the correlation between radiomic features and immune cell infiltration was analyzed.Results Totally 863 radiomic features were extracted from each region of interest.The AUC of ROC curve of radiomic model and clinical-ra-diomic model were 0.728 and 0.739 respectively.The survival difference was significant between high and low risk group.Besides,high risk group had significantly higher ESTIMATE score and high risk radiomic features tended to be related to higher regulatory T cell infiltration and lower dendritic cell and active CD4 memory T cell infiltration.Conclusion Con-siderably high predicting performance could be achieved by postoperative prognostic prediction of lung adenocarcinoma pa-tients using radiomic risk score.High risk group could have lower tumor purity,the ratio of tumor cells to stromal and im-mune cells,and had higher inhibitory immune cell infiltration.
NETL
NSTL
万方数据
