Study On Texture Analysis Of Dynamic Contrast-enhanced Magnetic Resonance Images To Distinguish Amplification Status Of HER2 2+ Breast Cancer

Objective: To explore the feasibility of texture features in identifying the amplification status of breast cancer patients with human epidermal growth factor receptor type 2(HER2)2+ based on dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI).Methods: We retrospectively included 92 breast cancer patients who had a HER2 score of 2+ tested by immunohistochemistry(IHC)and their HER2 amplification status were confirmed by fluorescence in situ hybridization(FISH).Among them,52 were HER2 2+positive and 40 were HER2 2+ negative.All of them underwent preoperative breast DCE-MRI.Export the precontrast,postcontrast,and subtraction images of the largest lesion diameter from the picture archiving and communication system(PACS).Region of interest(ROI)was semiautomatically delineated on the subtraction image,and map it to the corresponding position of the precontrast and postcontrast images using MATLAB2018 a.488 texture features were respectively extracted from the three images.The independent sample t-test or Mann-Whitney U test was performed to identify statistically significant features between different HER2 2+ amplification groups.Least absolute shrinkage and selection operator(LASSO)was used to search for the optimal feature subsets.Three machine learning classifiers,logistic regression analysis(LRA),quadratic discriminant analysis(QDA),and support vector machine(SVM),were used to establish the classification models.Classification performance was evaluated by receiver operating characteristic(ROC)analysis.Results: The area under the ROC curve(AUC)of the LRA model,SVM model and QDA model established by the texture features extracted from the subtraction images were 0.884,0.890 and 0.831,respectively.The AUC of the LRA model,SVM model and QDA model established by the texture features extracted from the precontrast images were 0.623,0.672 and 0.568,respectively.The AUC of the LRA model,SVM model and QDA model based on the texture features extracted from the postcontrast images were0.733,0.736 and 0.726,respectively.The subtraction image texture features showed the best classification performance among the three images,and the LRA and the SVM model reached significantly better performance than the QDA model(P = 0.0227 and P =0.0088,respectively).Conclusion: The texture features extracted from DCE-MRI images had the ability to identify HER2 2+ amplification status of breast cancer,especially features extracted from the subtraction images were helpful for discriminating HER2 2+ amplification status.