We are proud to celebrate the acceptance of the research paper “Multiple Shortcut Pathways in Mammography-Based Breast Cancer Risk Prediction” at the FAIMI 2026 Workshop on Fairness of AI in Medical Imaging, held in conjunction with MICCAI 2026.

The paper is authored by Judy Ashmawy, Habiba Mohamed, and Maya Hussien, under the supervision of Professor Dr. Tamer Basha and Eng. Alaa Melek, Assistant Lecturer at SBME. The work originated from the students’ graduation project and has now progressed to publication at an international scientific venue.

The research investigates an important challenge in AI-based breast cancer risk prediction from mammography: whether highly performing deep learning models are making predictions based on clinically meaningful imaging features or relying on unintended shortcut signals. Using techniques including occlusion analysis, background perturbation experiments, and representation analysis, the study identifies multiple shortcut pathways that can influence model predictions, highlighting the importance of interpretability, fairness, and clinically reliable AI in medical imaging.

FAIMI – Fairness of AI in Medical Imaging is a specialized workshop held in conjunction with MICCAI, the flagship international conference in medical image computing and computer-assisted intervention. FAIMI brings together researchers working to understand and mitigate bias, fairness, robustness, and generalizability challenges in medical AI, with the goal of supporting safer and more equitable translation of artificial intelligence into clinical practice.

The Systems and Biomedical Engineering Department proudly congratulates the research team on this achievement and wishes them continued success as they present their work to the international medical imaging research community at MICCAI 2026 in Strasbourg, France.

Read the paper:

Multiple Shortcut Pathways in Mammography-Based Breast Cancer Risk Prediction