Jack Breen

Research Topic: Developing Artificial Intelligence Solutions in Ovarian Cancer Diagnosis

Supervisors: Dr Nishant Ravikumar, Dr Nic Orsi, Dr Kieran Zucker, Prof Geoff Hall

About Jack: Jack received an integrated bachelors and masters degree in mathematics (MMath) at the University of Nottingham, with a focus on statistics and computational methods

Project description: This project investigates explainable, computationally efficient artificial-intelligence methods for histological sub-typing of ovarian cancer—an under-served yet clinically critical task—by first surveying the literature to reveal that previous work has been confined to small, homogeneous datasets with limited clinical focus, and then conducting the most comprehensive evaluation to date using the largest training cohort and rigorous validation protocols (cross-validation, hold-out testing, external cohorts, bootstrapping and hypothesis testing) built on an attention-based multiple-instance-learning framework with an ImageNet-pre-trained ResNet-50 backbone; to address the hardware constraints typical of pathology laboratories, the study introduces an active-tissue-sampling strategy that slashes inference costs while preserving diagnostic accuracy, demonstrates that 10× magnification offers the optimal trade-off between cellular detail and contextual histoarchitecture versus the de-facto 40× standard, benchmarks 14 emerging histopathology foundation models—confirming significant performance gains over traditional feature extractors—and proposes a multi-resolution patch-graph network to capture spatial context, ultimately achieving balanced accuracies of 88 %, 99 % and 77 % across three validation sets (compared with 66 %, 69 % and 52 % for the baseline and 74–91 % inter-pathologist concordance), thereby establishing a robust evidence base that such AI systems can attain clinical utility and motivating future work on real-world deployment and impact assessment.