Breast cancer is the most common malignancy affecting women worldwide and the fifth leading cause of cancer death. Survival of breast cancer patients in the past few decades has mostly improved through screening, especially if tumours are diagnosed at early stages. A comprehensive AI-assisted risk prediction model with an improved discriminatory power to classify women into clinically meaningful risk groups can inform the screening system in a complementary manner to the radiologists.
In the project, the candidate will help develop an Explainable AI-assisted digital breast cancer screening platform for high-throughput automated Mammogram analysis. The research will use the OPTIMAM database (accessible free of cost from Royal Surrey County Hospital NHS Foundation Trust) to access the health records and the associated mammograms from healthy and other interval breast cancer subgroups. This project will aim to develop disease detection and segmentation models using state-of-the-art computer vision and deep learning techniques such as Mask RCNN, and Neural Ordinary Differentiation Equation networks to predict breast density and breast cancer risk trajectory from mammograms along with associated health records. With this large and routinely collected mammogram data, we can unlock pertinent information and can capture complementary explanatory factors for various stages of cancer. Hence, the automated analysis techniques developed during this project have the potential to drive improvements in patient care.
This project will be conducted within the School of Computing Sciences under the supervision of Dr Tahmina Zebin at the University of East Anglia. There will be some essential collaboration with colleagues from the Royal Surrey County Hospital NHS Foundation Trust and Norfolk and Norwich University Hospital at various stages of the development. The project will provide the associated researcher with an opportunity to gain experience in rapidly expanding fields of computer science that include deep machine learning, risk prediction modelling, and digital health.