Fig. 2

The workflow of this study. This study includes image segmentation, feature extraction, feature selection, feature analysis and model validation. The ROIs of breast cancer was segmented and then 55 quantitative radiomics features were extracted from the segmentation images of individual patients. The least absolute shrinkage and selection operator was used for feature selection and signature construction. The performance of radiomics signature was evaluated by area the receiver operating characteristic curves. Finally, a nomogram was plotted based on the radiomics signature and clinical model, and the predictive performance of the nomogram was validated by the calibration curves and decision curve analysis