Mammograms ViT — Dynamic AI for Breast Cancer Risk
Sequence-aware vision transformers for 5-year breast cancer risk prediction from longitudinal mammography (CRUK ACED Pathway Award).
A two-year £190k CRUK International Alliance for Cancer Early Detection (ACED) Pathway Award, where I am the sole Principal Investigator (May 2026 – Present).
We are developing sequence-aware deep learning models — modern vision architectures and transformers — that consume longitudinal mammography to predict 5-year breast cancer risk and enable Dynamic Risk-Stratified Screening: stratifying women into high, medium, and low-risk groups to reduce over-screening of low-risk individuals while improving detection of interval cancers in higher-risk groups.
Co-mentors: Prof. Nora Pashayan and Prof. Sue Astley. We collaborate closely with Prof. Fiona Gilbert’s group in the Department of Radiology to ensure the methods remain clinically grounded, and engage with the wider ACED network spanning Cambridge, UCL, Stanford, Manchester, and Oregon.