Radiomics Of Oncotype DX Recurrence Score For Predicting Of Neoadjuvant Chemotherapy Responses And Survivals In Breast Cancer

Breast cancer is a high incidence malignant tumor which seriously threatens women’s health.Due to its high heterogeneity,the results of patients may be quite different.This difference is reflected in the risk of recurrence,the benefit of adjuvant chemotherapy,the efficacy of neoadjuvant chemotherapy(NAC),and the long-term survival rate.The key to improve the cure rate of breast cancer is to realize the accurate differentiation of patients and to take comprehensive and individualized treatment plan.Therefore,it is necessary to study the markers(labels)that can provide effective reference information for clinical decision-making.In recent years,with the rapid development of radiomics,the advantages of medical imaging,such as non-invasive,widely used and quantificationally depicting tumor characteristics,have become increasingly prominent.The exploration of imaging markers of breast cancer has gradually become the focus of attention.At present,preliminary progress had been made.However,in the experimental results at this stage,there were not many image features with predictive value,and most studies were limited to a single prediction target,and the potential value of image features was not fully explored.When extracting image features,this study expanded the calculation range from space,time,category and other angles,trying to mine effective parameters from a large number of data.In addition,this study not only established the association between radiomics and breast cancer recurrence risk,but also explored the predictive effect of 21-gene image signature based on imaging features on the efficacy of NAC and prognosis.The specific research contents include:(1)Prediction of Oncotype DX RS of breast cancer based on radiomics: the image features were preliminarily selected by normal distribution test and unsupervised feature selection method.The remaining features were analyzed by univariate linear regression to evaluate the association between single image features and RS.Then,multivariate linear regression analysis was used to evaluate the joint prediction effect of image features.The elastic network regression model was established by combining grid search and cross validation,and the evaluation indexes were calculated to evaluate the prediction performance of the model.(2)The application of 21-gene image signature in NAC efficacy prediction based on radiomics: elastic network regression model established the association between image features and RS,and calculated 21-gene image signature of neoadjuvant chemotherapy data set based on it.Statistical analysis was used to evaluate its association with chemotherapy efficacy.Objective to evaluate the value of 21-gene image signature in predicting the efficacy of NAC by comparing the performance of the logistic regression classification model established by other imaging features and clinicopathological features.(3)Application of 21-gene image signature in survival analysis based on radiomics: elastic network regression model established the association between image features and RS,and calculated 21-gene image signature of survival analysis data set based on it.The association of 21-gene image signature with recurrence-free survival(RFS)and overall survival(OS)were evaluated by using Kaplan Meier method and Cox proportional hazard regression model.Multivariate Cox analysis was performed with other imaging features and clinicopathological features to evaluate the independent correlation between 21-gene image signature and RFS and OS.In this paper,we studied the association between radiomics and recurrence risk of breast cancer,established 21-gene image signature based on image features,and explored its application in NAC efficacy prediction and survival analysis.The experimental results show that dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)was a promising imaging marker for predicting the recurrence risk and chemotherapy benefit of breast cancer.The 21-gene image signature based on image features has potential as a biomarker to predict the efficacy of NAC and differentiate the chemosensitivity of patients.The imaging marker of 21 gene in this study has the potential to be a marker for the treatment and prognosis of breast cancer and provide effective information for clinical decision-making.