Value Of CT-based Radiomics In Predicting The Efficacy Of Drug Therapy In Patients With HER-2 Positive Breast Cancer Liver Metastases

Objective:Breast cancer is the most common cancer in women,and liver is the third most common site of breast cancer metastasis,especially breast cancer with positive HER-2 is the most prone to liver metastasis.In the treatment of advanced breast cancer,accurate prediction of efficacy is the key to develop effective treatment plan.However,due to various factors such as tumor heterogeneity,even breast cancer patients with the same molecular subtype may respond differently to treatment.Therefore,this study intends to use the radiomics analysis to explore its value in predicating efficacy of drug therapy in patients with HER-2 positive breast cancer liver metastases.Methods:A total of 83 patients with HER-2 positive breast cancer liver metastases who received combined drug therapy including anti-HER2 therapy in three different hospitals from January 2011 to November 2021 were retrospectively included.Query the medical record information system and imaging data to obtain basic information about the patient’s age,hormone receptor and HER-2 expression status,general physical condition(ZPS score),medication use,information about recurrence or metastasis,and the progression free survival(PFS)with a combination of drugs including anti-HER2 therapy,and information on enhanced CT images of the abdomen after liver metastasis and before the use of anti-HER2 drugs.The enrolled patients were divided into good prognosis and poor prognosis groups according to uniform criteria,and the patients were stratified and randomized into training and validation sets in a 7:3 ratio.The largest cross-section of the largest liver metastases in CT images was used as the region of interest(ROI)for radiomics feature extraction,and the variance threshold,SelectKBest and Least absolute shrinkage and selection operator(LASSO)regression models were used to select effective radiomics features in three different periods of enhanced CT,respectively.Using six classifiers:k-nearest neighbor(KNN),support vector machine(SVM),eXtreme Gradient Boosting(XGBoost),random forest(RF),logistic regression(LR)and decision tree(DT),to construct the radiomics models.Accuracy(score)matrix,receiver operating characteristic(ROC)curve,precision,recall,F1 score,and support were used to evaluate the prediction performance of different classifiers.Results:1.In the portal venous phase of enhanced CT,SVM classifier had the best performance,and four significant radiomics features were selected.The AUC value of the prediction model was 0.865 in the verification set.2.In the arterial phase,XGBoost classifier had the best performance.The AUC value of the model composed of four selected radiomics features in the verification set was 0.601.3.In the delay period,the LR classifier had the best performance.The AUC value of the prediction model composed of seven radiomics features in the verification set was 0.628Conclusion:1.Radiomics analysis have high diagnostic performance in predicting the effectiveness of drugs including anti-HER2 therapy in patients with breast cancer liver metastases.2.The liver metastases in the portal venous phase of enhanced CT are the most suitable for the prediction of efficacy.3.The development of radiomics models is helpful to guide the treatment of advanced HER-2 positive breast cancer patients,avoid ineffective treatment as much as possible,and help promote the development of personalized and precise treatment.