HRCT Diagnosis And Radiomics On The Invasiveness Of Pure Ground-glass Nodules In Pulmonary Adenocarcinoma

Objective:This research retrospectively analyzed the CT features of pure ground glass nodules in pulmonary adenocarcinoma on HRCT at first,then distinguished the invasiveness of pGGN according to the pathology results,and explored whether HRCT combined with conventional imaging diagnosis could predict the invasiveness of pGGN in pulmonary adenocarcinoma.Secondly,analyzed the original HRCT imaging data by radiomics and dug the quantitative characteristics of medical image data in depth.This part wanted to explore whether radiomics could predict the invasiveness of pGGN,and whether the its diagnostic value is superior to conventional imaging diagnosis method,so as to find a more accurate method for pGGN lesions,in order to assist clinical diagnosis and choose rational operate methods of pGGN in pulmonary adenocarcinoma.Methods:The retrospective analysis of 65 cases with pGGN which underwent HRCT examination and postoperative pathological results confirmed pulmonary adenocarcinoma from January 2015 to October 2018 in the Second Hospital of JiLin University were enrolled in this study.According to the presence of invasive component,65 cases were divided into pre-invasive group(38 cases)and invasive group(27 cases).The pre-invasive group included 7 cases of atypical adenomatous hyperplasia(AAH)and 31 cases of adenocarcinoma in situ(AIS).The invasive group included 20 cases of minimally invasive adenocarcinoma(MIA)and 7 cases of invasive adenocarcinoma(IAC).All nodules in the two groups were analyzed from qualitative features and quantitative features and gained the value of conventional imaging diagnosis on invasiveness of pGGN pulmonary adenocarcinoma.Then putted all DICOM images of the 65 cases into the A.K.software(Artificial Intelligent Kit,GE Healthcare)to automatically obtain the radiomics parameters based on the first-order histogram features,high-order texture features and morphological features.A.K.software reduced the dimentions of all features obtained,and established a radiomics prediction model.ROI imaging data of test set were brought into the model and carried out statistical analysis by A.K.software automatically.Results:In conventional imaging diagnosis method,the features between pre-invasive and invasive group were analyzed qualitatively and quantitatively,and the results showed that Lobulation(P=0.000),Speculation(P=0.008),Bubble-like lucency(p=0.047)and Vascular cluster(P=0.001)of qualitative features were significantly different between the two groups.The nodule size(P=0.001)and average CT value(P=0.000)were statistically significant,the optimal threshold value was 12.7 mm and-544.50 HU and the AUC was 0.740 and 0.784 respectively.A.K.software automatically extracted 385 radiomics features from train set.LASSO and Logistic regressive were used to eliminate redundancy,and finally obtained two parameters “Entropy” and “Sphericity”.We used two parameters to establish the prediction model then brought test set into this model,and the ROC curve was drawn after statistical analysis.In this study,the diagnostic AUC value of the radiomics model for pulmonary adenocarcinoma pGGN in HRCT was 0.824.Conclusion:HRCT images of 65 cases pGGN in pulmonary adenocarcinomas with diameter less than 30 mm were qualitatively and quantitatively analyzed by conventional imaging diagnosis methods.The results showed that Lobulation,Speculation,Bubble-like lucency,Vascular cluster,nodule size and average CT value were significant in differentiating pre-invasive and invasive lesions.When nodule diameter was greater than 12.7 mm or average CT value was greater than-544.50 HU could be considered as an invasive lesion.Using the radiomics method to analyz the 65 lesions mentioned above and established an analysis model to distinguished the invasiveness of pGGN.The results showed that this model can not only quantitatively judge whether there is invasiveness in pGNN,but also has higher diagnostic value than conventional imaging diagnosis methods.These studies provide a new method for noninvasive preoperative diagnosis of pGGN invasiveness in pulmonary adenocarcinoma,and provide a new idea for clinical application of radiomics.