CT Signs Combined With Lung-RADS Grading In Identification Of SPN And Predicting EGFR Mutation In Non-small Cell Lung Cancer

Part one: A Study of Conventional CT Signs Combined with Lung-RADS Grading for the Identification of Benign and Malignant Solitary Pulmonary NodulesObjective: This study is intended to evaluate the value of imaging signs of conventional plain CT,Lung-RADS system grading criteria and CT imaging signs combined with the Lung-RADS system grading criteria for the differentiation of benign and malignant solitary pulmonary nodules（SPN）.Methods: A total of 164 patients with 164 lesions of SPN diagnosed with acute pathological by radical surgery or aspiration biopsy in our hospital from October 2020 to October 2021 were analyzed retrospectively.Through multiplanar reconstruction,all patients’ lesions were scored by Lung-RADS grading and the signs of lesions on CT images were interpreted.By using chi-square test,the differences of qualitative imaging morphological features between groups were analyzed.To use the receiver operating characteristic（ROC）curve calculate the area under curve（AUC）,specificity and sensitivity,then analyzing the relationship among AUC by Delong test which was used to compare the magnitude of the differential diagnosis efficiency for benign and malignant SPN between CT sign,Lung-RADS grading system and CT sign combined with Lung-RADS grading system.Results:（1）Between benign and malignant SPN groups,the differences were statistically significant in nodule morphology,pleural indentation sign,lobulation sign and spiculation sign（P<0.05）.The independent risk factors for malignant SPN were spiculation sign and irregular nodules.（2）The cutoff value,sensitivity,specificity and AUC of the efficiency to diagnose benign and malignant SPN with multi-CT signs were0.4429,92.04%,45.10% and 0.749（95%CI 0.676-0.814）.The cutoff value,sensitivity,specificity and AUC of the efficiency to diagnose benign and malignant SPN with Lung-RADS grading system were 4,66.37%,86.27% and 0.765（95%CI 0.693-0.828）.The cutoff value,sensitivity,specificity and AUC of the efficiency to diagnose benign and malignant SPN with CT sign combined with Lung-RADS grading system were0.39633,92.92%,45.10% and 0.798（95%CI 0.728-0.856）.The efficiency of CT signs combined with Lung-rads system in differentiating benign and malignant SPN was significantly higher than CT signs singly,but there was no significant difference compared with Lung-rads system.Conclusion: Both Lung-rads grading and conventional CT signs have certain diagnostic value for benign and malignant SPN,and the efficiency can be improved after their combination.Part two: Predicting the EGFR Mutation of Non-small Cell Lung Cancer Based on CT Signs and Radiomics.Objective: This study is based on the results of epidermal growth factor receptor（EGFR）gene detection as the reference standard,intended to predict the EGFR mutation status for preoperative non-small cell lung cancer（NSCLC）patients noninvasively based on CT signs and radiomics model.Methods: All CT images and clinical data of 94 patients with lung adenocarcinoma that were diagnosed by radical lung cancer surgery or aspiration biopsy were gathered retrospectively from January 2018 to December 2021.The imaging signs were evaluated on the chest plain CT images.The region of interest（ROI）was divided manually in three dimensions by using ITK-SNAP software.The CT radiomics features were extract by 3D-Slicer software,then using Spearman correlation and least absolute shrinkage and selection operator（LASSO）to decrease the dimension of the features twice through IPM software.Constructing radiomics model by the features with the most predictive value,then calculating the area under curve（AUC）,specificity,sensitivity,negative and positive predictive value by using receiver operating characteristic（ROC）curve to evaluate its efficiency in predicting EGFR mutation.The goodness of fit of radiomics model was appraised by calibration curve and the decision curve analysis（DCA）was used to appraise the clinical value of the model.Results:（1）There were significant differences in gender,smoking history and Ki-67 between EGFR mutation group and wild group by univariate analysis.（2）There were six features（mean,energy,skewness,cluster shadow,large dependence low gray level emphasis,gray level non-uniformity normalized）used to construct the model,which AUC was 0.863（95%CI 0.787-0.931）in the training samples and 0.740（95%CI 0.576-0.887）in the testing samples.The sensitivity,specificity,positive predictive value and negative predictive value of training samples were 0.886,0.733,0.795,0.846 and were 0.765,0.583,0.722,0.636 of testing samples.The DCA showed there was certain clinical practicability of the model.Conclusion: There were correlation between gender,smoking history,ki-67 with EGFR mutation in NSCLC.The radiomics model based on plain CT images had certain predictive value for EGFR mutation in pulmonary adenocarcinoma.