Study Of Imaging Features And Radiomics Model Based On Energy Spectrum CT To Predict Early Lung Adenocarcinoma In Filtration

Objective:The purpose of this study was to explore the value of manual imaging model CT energy spectrum and imaging group model based on energy spectrum CT in predicting early lung adenocarcinoma(cTINOM0)infiltration before operation.Materials and methods:Prospectively collected 196 patients with early-stage Lung Adenocarcinoma who underwent surgical treatment,pre-operative standard dual-energy scan,and post-operative pathology at the Third Affiliated Hospital of Kunming Medical University from September 2019 to November 2020,there were 50 cases of preinvasive lesions and 146 cases of invasive lesions,the pre-invasive Lesions and invasive lesions were divided into training set and independent verification set according to the ratio of 8:2.ROI 1 and ROI 2 are drawn on the lung window by the software of radiomics,which can draw the ROI 1 and ROI 2 at the same time,and extract the features of radiomics using Lasso to preserve the features closely related to lung Adenocarcinoma invasion and to establish an radiomics model;manual features,including qualitative and quantitative features,were extracted on the same sequence,after screening the manual features and clinical features by single factor and multi-factor analysis,a hybrid clinical manual imaging model was established,and then the hybrid model was established by combining the imaging features with the manual imaging features and clinical features,the diagnostic efficacy of the three models was compared using postoperative pathological results as the gold standard.Results:(1)63 imaging omics features were retained by single-factor analysis,and 10 imaging omics characteristics closely related to early lung adenocarcinoma infiltration were finally selected by LASSO dimension reduction,including 5 from the lesion itself,and 5 were from the tumor microenvironment;the AUC values in the training set and validation set were 0.86 and 0.87,respectively.2)The results of univariate analysis showed that sex,age,active smoking and place of origin were high incidence of lung cancer,the maximum diameter of the lesion,the CT value of plain scan and the CT value of single energy imaging in enhanced arterial phase,the type of nodule,the shape of nodule and pleural traction depression were statistically significant between the two groups;The values of nodule type,active smoking,pleural traction and depression,enhanced arterial phase scan 40/90/150 keV single-energy imaging CT were included in the mixed clinical manual imaging model,AUC values of the model in training set and validation set are 0.82 and 0.76 respectively;(3)The AUC values of the mixed model was 0.88 in both the training set and the validation set,and only the imaging score was an independent predictor of early lung adenocarcinoma invasion.Conclusions:1.The qualitative and quantitative characteristics of early lung Adenocarcinoma were analyzed based on the conventional imaging diagnostic method of energy spectrum,the types of nodules,pleural retraction and indentation,enhancement arterial scanning 40/90/150 Kev single-energy CT scan were found to be significant in differentiating pre-invasive from invasive lesions.2.The characteristics of the peritumoral microenvironment extracted by imaging were closely related to the invasion of early lung Adenocarcinoma.3.The Radiomics model based on energy spectrum CT has a good ability to predict the invasion of early lung Adenocarcinoma,which can be used as a reference for the preoperative selection of patients with early lung Adenocarcinoma,the combination of clinical features and hand-made imaging features is more effective in prediction.