Ct Radiomics Features Of The Peritumoral Microenvironment Predicting The Invasion In Pure Ground Glass Nodule Lung Adenocarcinoma

Objective:To explore the diagnostic value of CT radiomics features of peritumoral microenvironment in the invasion of pure ground glass nodule lung adenocarcinoma.Method:A retrospective analysis of 100 pure ground glass nodules from patients who underwent surgical treatment at the First Hospital of Jilin University from January 2017 to May 2019.Ground glass nodules were randomly divided into train set,verification set,and test set according to the ratio of 6: 2: 2.Lung tissues which were around the pure ground glass nodules for the distance of 5mm were selected as the microenvironment,and 1389-dimensional radiomics features were extracted from it.Applying the least absolute shrinkage and selection operator method to reduce the dimensions to obtain 16-dimensional features,and used the five-fold cross-validation method to establish six types of radiomics signature models which were named random forest classier,gradient boosting decision tree,multi-layer perceptron,support vector machine,logistic regression and stacker classifer.The better efficiency logistic regression model was chosen to combine with qualitative and quantitative features to establish a new one.The postoperative pathological results were used as the gold standard to evaluate the diagnostic efficacy of the two models.Results:There were no statistically significant differences in age,gender,location,lesion shape,vacuole sign,pleural sign,and lesion density these qualitative and quantitative features.Tumor boundary,diameter and volume which were related to the invasion of pure ground glass nodule.The accuracy of the radiomics signature model was 0.86±0.07,the mean AUC of train set was 0.96±0.01,the mean AUC of verification set was 0.94±0.06,the AUC of test set was 0.917.The accuracy of the radiomics signature combined with qualitative and quantitative features model was 0.86 ± 0.05,the mean AUC of train set was 0.95±0.02,the mean AUC of verification set was 0.94 ±0.07,the AUC of test set was 0.938.Conclusion:Logistic regression model based on CT radiomics features of peritumoral microenvironment,which had a good ability to predict the invasion of pure ground glass nodule,could provide a reference for clinical diagnosis and treatment.The predictive performance of radiomics signature combined with qualitative and quantitative features model was better.