AI-Powered Breast Cancer Screening and Molecular Profiling

Breast cancer is a leading cause of mortality, with early and precise diagnosis critical for effective treatment. Current diagnostic workflows rely on mammography for initial screening, but confirmation of receptor status (HER2, ER, PR) requires biopsy and immunohistochemical analysis, which can be invasive, costly, and time-consuming.

This project seeks to develop an AI-based decision support system that utilises deep learning to predict breast cancer receptor status directly from mammograms. The approach eliminates the need for manual segmentation, leveraging convolutional neural networks (CNNs) and attention-based architectures to enhance diagnostic accuracy. The system will be designed for clinical and commercial deployment, integrating with existing digital health infrastructures to support radiologists and oncologists in treatment planning.

Objectives:

• Develop and validate a deep learning model to predict HER2, ER, and PR expression from mammography images.

• Integrate AI-driven computer-aided diagnosis (CAD) to improve breast cancer detection.

• Provide a non-invasive alternative to biopsy-based molecular profiling.

• Ensure the system is scalable for use in hospitals, screening programmes, and telemedicine platforms.

• Conduct multi-centre validation to support regulatory approval and clinical adoption.

Expected Impact:

• Clinical Application: Enhanced breast cancer detection and treatment planning.

• Patient Benefit: Faster, less invasive diagnosis, reducing the need for biopsies.

• Commercial Potential: A deployable AI solution for medical imaging providers, health technology firms, and research institutions.