Cross-Attention Multimodal Fusion for Breast Cancer Diagnosis: Integrating Mammography and Clinical Data with Explainability
This study developed a multimodal deep network to diagnose breast cancer by combining mammography images with clinical data. The model uses cross-attention fusion between image features and structured clinical features, achieving high performance metrics like an AUC-ROC of 0.98. Explainable AI techniques (Grad-CAM, SHAP) were used to visualize important image regions and feature contributions, validating the model's predictions.