ObjectiveTo investigate the application value of CT radiomics in the pathological grading of non-small cell lung cancer (NSCLC).
MethodsThe data of 67 patients with pathologically confirmed NSCLC were analyzed retrospectively.Among them, there were 7 cases of grade Ⅰ, 39 cases of grade Ⅱ and 21 cases of grade Ⅲ.According to the degree of differentiation, atypia and number of mitotic images of NSCLC, the patients of grade Ⅰ and Ⅱ were classified into the low-grade group (46 cases) and the patients of grade Ⅲ were classified into the high-grade group (21 cases).Patients were randomly divided into the training group (56 cases) and the verification group (11 cases) according to the ratio of 5∶1.The maximum diameter of the lesions was selected to delineate the area of interest (ROI), and the prediction model was established.By drawing the receiver operating characteristic curve (ROC) and calculating the area under ROC curve (AUC), the application value of radiomics features in the pathological grading of NSCLC was evaluated.
ResultsA total of 1878 radiomics features were extracted from the CT radiomics model.SVM was used to evaluate the importance of the features, and finally 20 feature dimensions were obtained.In the training group, the AUC value was 0.851, and the accuracy was 80.35%.In the test group, the AUC was 0.833 and the accuracy was 90.90%.
ConclusionsCT radiomics can predict the pathological grade of preoperative NSCLC by analyzing various image features.
DownLoad: