Radiomic-based Quantitative CT Analysis Of Pure Ground-glass Nodules For The Identification Of The Invasiveness Of Lung Adenocarcinoma

Objectives:To investigate the performance of radiomic-based quantitative analysis on CT images in identifying invasiveness of lung adenocarcinoma manifesting as pure ground glass nodules(p GGNs).Methods:275 lung adenocarcinoma cases,with a total of 322 p GGNs resected surgically and confirmed pathologically,from January 2015 to October 2017 were enrolled in this retrospective study.Radiomic feature extraction was performed using Pyradiomics with semi-automatically segmented tumor regions on CT scans which were contoured with an in-house plugin for 3D-Slicer.Random forest(RF)and Support Vector Machine(SVM)were used for feature selection and predictive model building.The predictive performance of each model was evaluated through the receiver operating characteristic curve(ROC curve).Results:Among 322 nodules,150(46.6%)were Adenocarcinoma in situ(AIS)and minimally invasive adenocarcinoma(MIA)and 172(53.4%)were invasive adenocarcinoma(IVA).All nodules were split into training and test cohort randomly with a ratio of 4:1 to establish predictive models.Three different predictive models containing conventional,radiomics and combined models were created using training cohort.The area under the curve(AUC)values in test cohort were 0.866(0.778?0.954)for combined model with 79.69%,88.24%,70.00%,84.00% and 76.92% for accuracy,sensitivity,specificity,negative predictive value(NPV)and positive predictive value(PPV)respectively.Radiologists could improve their diagnostic accuracy for the discrimination of p GGN-like IVA and AIS/MIA significantly with the aid of radiomics predictive model built in our research.Conclusion:The predictive models created in our study showed significant predictive power with high accuracy and sensitivity,which provided a non-invasive,convenient,economic and repeatable way for the identification of IVA from AIS/MIA representing as pGGNs.