Interpretable breast cancer identification by multi-view synergistic feature fusion interaction in dense breast tissue

Guowei Dai, Chaoyu Wang, Duwei Dai, Qingfeng Tang, Yi Zhang, Hu Chen

Abstract: Highlights•Novel EAGANet integrates EdgeNet and LWM-Mamba for dense cancer detection.•Dynamic weighting via LWM-Mamba allows precise detection in complex mammograms.•Breast density-aware framework achieves 97.91% accuracy via dual-stream fusion.•Multi-level defense strategy (REF, CEMSP, CNFM) ensures model robustness.•Hierarchical interpretability system enables clinically relevant visualization.