Breast Cancer Prediction: BI-RADS 4/5 and Radiomics

Radiomics, an advanced imaging technology, enables quantitative analysis of tumor texture and morphology from ultrasound (US) images. This study introduces a nomogram that combines radiomics and Breast Imaging Reporting and Data System (BI-RADS) classifications to predict breast cancer in BI-RADS US category 4 or 5 lesions. Based on 315 pathologically confirmed breast lesions, the dataset was split into a training group and a validation group. Radiomics scores were derived from US images by extracting nine key imaging features that are beyond the capability of visual interpretation. A predictive nomogram was developed by integrating these radiomics scores with BI-RADS categories, which classify lesions into subcategories (4A, 4B, 4C, and 5) based on malignancy probability. The resulting model demonstrated exceptional predictive accuracy, achieving an area under the receiver operating characteristic curve (AUC) of 0.928 in the validation group, outperforming either radiomics scores or BI-RADS categories alone. This integrated approach enhances diagnostic precision for distinguishing between malignant and benign breast lesions. The nomogram also showed excellent calibration and strong clinical applicability, offering a reliable tool for guiding biopsy decisions. Focused on BI-RADS US category 4 and 5 lesions, which have a wide range of malignancy probabilities, the model provides a more precise risk stratification. By combining radiomics with BI-RADS categories, the study represents a significant advancement in breast cancer diagnostics, potentially reducing unnecessary biopsies and improving patient care. This innovative approach demonstrates how radiomics can complement traditional imaging methods, providing a new standard for early and accurate breast cancer detection in clinical practice.