Study On The Application Of Radiomics Based On Magnetic Resonance Imaging In Evaluation Of Biological Behavior Of Breast Cancer

Objective To investigate the feasibility and application value of radiomics model based on fat-pressure T2 WI image and DCE-MRI image to predict the expression of biological indexes ER,PR,Her-2 and Ki-67 in breast cancer.Methods In this study,122 cases of breast cancer patients confirmed by postoperative pathology were retrospectively analyzed,and preoperative MRI scan(plain scan + enhanced scan)was performed.The software ITK-SNAP was used to manually draw ROI on fat-pressure T2 WI and DCE-MRI images,respectively.Artificial Intelligence Kit(GE Healthcare,Life Science,China)was used to extract and analyze the radiomics features.Using immunohistochemical staining to detect the expression levels of ER,PR,Her-2 and Ki-67 in tissues as reference criteria,single-factor logistic regression,correlation analysis and lasso algorithm(Least Absolute Shrinkage and Selection Operator,LASSO)was used for feature screening and dimensionality reduction.Logistic regression algorithm was used to build models respectively.Receiver operating characteristic(ROC)curves were drawn to evaluate the effectiveness of the model in predicting the expression level of biological indicators.Results According to the expression levels of ER,PR,Her-2 and Ki-67,7,7,6 and 3 key features were selected for modeling on fat-pressure T2 WI images,respectively,and 3,5,5and 6 key features were selected for modeling on DCE-MRI images.The AUC values of ER,PR,Her-2,Ki-67 prediction models based on fat-pressure T2 WI images were 0.783(95%confidence interval: 0.700-0.861),0.760(95% confidence interval :0.686-0.829),and 0.746,respectively(95% confidence interval :0.654-0.828),0.754(95% confidence interval :0.676-0.829).The AUC values of ER,PR,Her-2 and Ki-67 prediction models based on DCE-MRI images were 0.700(95% interval :0.596-0.795)and 0.729(95% interval :0.652-0.802),0.695(95% confidence interval :0.611-0.775),0.779(95% confidence interval :0.704-0.847).Conclusion(1)Radiomics model based on MRI images(fat-pressure T2 WI and DCE-MRI)can predict the expression of biological indicators(ER,PR,Her-2,Ki-67)of breast cancer,which can provide quantitative parameters about the biological behavior of breast cancer in clinic and indirectly evaluate the biological behavior of breast cancer in vivo.So as to provide the basis for the formulation of personalized treatment plan for patients.(2)Among the established models,the Logistic regression model based on fat-pressure T2 WI image to predict the expression of ER has the highest efficiency,and the Logistic regression model based on DCE-MRI image to predict the expression of Ki-67 has the highest efficiency.