关键词: 乳腺癌, 影像组学, 新辅助治疗, 乳腺X线摄影, 磁共振成像

Abstract: Objective To investigate the value of machine learning models based on mammography and dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)in predicting pathological complete response(pCR)in breast cancer patients after neoadjuvant therapy(NAT). Methods A retrospective analysis of 396 breast cancer patients who underwent NAT followed by surgery from August 2016 to July 2023 at the Affiliated Hospital of Qingdao University(Institution 1)and Yantai Yuhuangding Hospital(Institution 2)was performed. A total of 320 patients from Institution 1 were randomly divided into a training set and a validation set in a ratio of 7∶3, and 76 patients from Institution 2 served as an independent external validation set. Regions of interest(ROI)were delineated on pre-NAT mammography and DCE-MRI images, followed by feature extraction and feature selection. The radiomics model was constructed using the support vector machine(SVM)algorithm. Clinical features underwent univariate and multivariate analyses, and statistically significant independent predictors were used to construct the clinical model. The comprehensive model integrating the radiomics signature and clinical predictors was constructed using the SVM algorithm. The performance of the models was evaluated using the area under the receiver operating characteristic(AUC)curve, accuracy, sensitivity, specificity, and F1-score. The calibration efficiency of the predictive models was evaluated by drawing calibration curves and decision curve analysis(DCA)was performed to evaluate the clinical utility of the predictive models. Results The combined radiomics model demonstrated better predictive performance than the clinical model, mammography radiomics model and MRI radiomics model, with AUCs of 0.899, 0.850, and 0.765 in the training, validation, and external validation sets, respectively. The comprehensive model showed the best predictive performance, with AUCs of 0.918, 0.856, and 0.795 in the training, validation, and external validation sets, respectively. The comprehensive model exhibited good calibration ability and clinical benefit. The Delong test showed statistically significant difference between the clinical model and the comprehensive model(P<0.05). Conclusion Machine learning models based on mammography and DCE-MRI effectively predict pCR in breast cancer patients after NAT, and demonstrate preferable predictive performance.

Key words: Breast cancer, Radiomics, Neoadjuvant therapy, Mammography, Magnetic resonance imaging