The Initial Research On The Image Texture Features Of Different Pathological Types Of Lung Cancer Brain Metastases And Preliminary Diagnosis Of Combined NSE

Objective:By analyzing the conventional MR image texture features of the tumor solid components and peritumoral edema from different pathological types of lung cancer brain metastases,the differences in texture characteristics of lung cancer brain metastases with different pathological types were analyzed.Evaluating the diagnostic performance of texture features combined with NSE,in order to provide important information to personalized treatment of lung cancer brain metastases.Materials and Methods:This study included a total of 74 patients with metastatic tumors confirmed by pathology of lung cancer brain metastases.General clinical data were collected including gender,age,and Serum tumor markers?CEA and NSE?as well as conventional MR images.The MR images were processed by ITK-SNAP software and A.K.software.After extracting and screening the texture feature,the model was established.Evaluating the diagnostic performance of the established model.The optimal model was selected to construct the radiomics score.Statistical analysis of the general clinical data of the non-small cell lung cancer and small cell lung cancer groups was performed using SPSS software,and the diagnostic efficacy of the NSE combined radiomics score was evaluated by MedCalc software.Result:By analyzing the texture features of conventional MRI images,only the enhanced T₁WI image texture features of solid tumor components are different between non-small cell lung cancer and small cell lung cancer brain metastases.The established four models including SVM?KNN?NB and Logistic evaluated well.The Logistic regression model performed best,with an AUC value of 0.84375,an accuracy of 0.79,a sensitivity of 0.75,and a specificity of 0.815.The diagnostic performance was improved when combined with the Serum tumor marker NSE,which was more stable.Conclusion:The solid tumor component enhancement T₁WI image texture features can be used to characterize the pathological types of non-small cell lung cancer and small cell lung cancer brain metastases,and the diagnostic performance is good.It can provide important information for the personalized treatment of lung cancer brain metastases.