Artificial intelligence-driven multimodal fusion for precision diagnosis and personalized management of breast cancer

Breast cancer is the most common malignancy among women worldwide, characterized by pronounced heterogeneity across molecular profiles, imaging phenotypes, and the tumor microenvironment. As precision oncology continues to advance, diagnostic strategies that rely predominantly on single-modality imaging or pathology face inherent limitations in early detection, individualized risk stratification, and timely recurrence assessment. Recent progress in artificial intelligence, particularly deep learning-based methods, has accelerated the development of multimodal fusion models that integrate radiomics, digital pathology, multiomics data, liquid biopsy biomarkers such as circulating tumor DNA, and clinical variables within unified computational frameworks. These integrative approaches enable the discovery of cross-modal associations and have demonstrated improved performance in diagnosis, molecular subtyping, treatment response prediction, and prognostic evaluation. In this review, we provide a comprehensive overview of the fundamental principles, methodological advances, and representative clinical applications of AI-driven multimodal fusion in breast cancer. We further discuss emerging directions, including digital twin-based modeling, dynamic monitoring of minimal residual disease, and multimodal large language models. Finally, we highlight current challenges related to data standardization, model interpretability, and multi-center validation, and outline future perspectives toward clinically translatable and robust intelligent systems.