Predicting The Invasiveness Of Pulmonary Adenocarcinomas In Pure Ground-glass Nodules Based On Radiomics Of Low-dose Computed Tomography

Objective:To develop a radiomic model based on low-dose computed tomography(LDCT)to predict the invasiveness of lung adenocarcinoma appearing as pure groundglass nodules(pGGNs)and compare its performance with traditional semanticquantitative model and intraoperative frozen section model.Materials and method:A total of 586 consecutive pathologically confirmed pGGNs from January 2019 to December 2022 were divided into a primary cohort[235 adenocarcinomas in situ/minimally invasive adenocarcinomas(AIS/MIAs)and 193 invasive adenocarcinomas(IACs)]and validation cohort(94 AIS/MIAs and 64 IACs)according to scans(Somatom Definition Flash and Brilliance iCT).Least absolute shrinkage and selection operator(LASSO)was used for feature selection.The radiomic radiomics,traditional quantitative-semantic model,and intraoperative frozen section model were built using multivariable logistic regression.The diagnostic performance was assessed by area under curve(AUC)of receiver operating characteristic curve,sensitivity,and specificity.The DeLong test was used to compare the AUCs among models in the primary and validation cohorts.Result:The AUCs of the radiomic model,traditional semantic-quantitative model,and intraoperative frozen section model were 0.940[95%confidence interval(CI):0.9130.960],0.920(95%CI:0.890-0.944),and 0.910(95%CI:0.878-0.935)in the primary cohort,and 0.905(95%CI:0.848-0.946),0.913(95%CI:0.858-0.952),and 0.911(95%CI:0.855-0.950)in the validation cohort.No significance of AUC was found among the radiomic model,traditional semantic-quantitative model,and intraoperative frozen section model in the primary or validation cohort(All P>0.05)Conclusion:The radiomic model and traditional semantic-quantitative model based on LDCT,with excellent predictive performance,can be preoperative and non-invasive biomarkers to assess the invasive risk of pGGNs in lung cancer screening.