Use Of A Radiomics Model To Predict Tumor Invasiveness Of Pulmonary Adenocarcinomas Appearing As Pulmonary Ground-Glass Opacity Nodules

Objective:Preoperatively identifying the classification of lung adenocarcinoma appearing as ground-glass nodules according to the imaging is of great importance for guiding patients to choose the best treatment strategy.However,there are still great challenges in the application of clinical imaging diagnosis.For this purpose,a radiomics model was developed to research it's value in predicting the tumor invasiveness of pulmonary ground-glass nodules on chest CT.Methods:A total of 570 patients with 599 GGNs[including 201 preinvasive lesions and 398 minimally invasive and invasive pulmonary adenocarcinomas(IPAs)]were enrolled.And the clinical features of the patients,the qualitative imaging features and the quantitative features based on the texture analysis of the nodules were also be collected.All cases were randomly divided into two groups,the training cohort with the test group was about 4.2 to 1.The primary cohort was consisted of 484 nodules,The validation cohort was comprised of 115 nodules.In the primary cohort,univariate,multivariate,and logistic regression analyses were used to selected the significant characteristics to construct a radiomics nomogram to predict the invasiveness of GGNs,and it was verified by the validation cohort.The average of randomly multiplicating test results was as final results,and the ROC curve was used to evaluate the prediction efficiency of the model.Results:In primary cohort,preinvasive lesions were distinguished from IPAs based on pure or mixed GGN(PM),lesion shape,lobulated border,and pleural retraction and 35 quantitative parameters(P<0.05)using univariate analysis.Multivariate analysis showed that PM,lobulated border,pleural retraction,age,and fractal dimension(FD)were significantly difference between preinvasive lesions and IPAs.After logistic regression analysis,PM and FD were used to develop a prediction nomogram.The area under curve(AUC)was 0.76[95%CI:0.71 to 0.80]in ROC curve for the primary cohort.The AUC was 0.79[95%CI:0.71 to 0.88]in the validation cohort.Conclusion:For GGNs,PM,lobulated border,pleural retraction,age,and FD were the features related with invasiveness of pulmonary adenocarcinomas.The radiomics model based upon PM and FD may discriminate the preinvasive lesions from IPA appearing as GGNs.