Interpretable breast cancer diagnosis using histopathology and lesion mask as domain concepts conditional simulation ultrasonography

Guowei Dai, Chaoyu Wang, Qingfeng Tang, Yi Zhang, Duwei Dai, Lang Qiao, Jiaojun Yan, Hu Chen

Abstract: Highlights•Novel SgmaFuse framework integrates histopathology concepts with ultrasound imaging diagnostics.•Spatially guided multi-level alignment mechanism coordinates cross-modal feature interactions.•Histological semantic activation vector learning enhances model interpretability and accuracy.•Visual attention maps and concept-level contributions provide structured diagnostic explanations.•The framework validates concept-based interpretable learning across medical domains.