This study presents a federated learning system for cancer prognosis using decentralized histopathological images. The framework preserves patient privacy while enabling robust model training across institutions, facilitating collaborative analytics without data centralization and ensuring high diagnostic accuracy in oncology workflows.

Lucas Henry Wainwright, Alexandra Juliet Forsythe, Benjamin Arthur Callister, Eleanor June Maddison, Peter Zachary Rowell

This research explores the use of transformer-based models to predict postoperative complications from heterogeneous EHR time series. By learning from sequential data patterns across patients, the system enhances early intervention capabilities and supports clinical decision-making for improved surgical outcomes.

Matilda Rose Halberd, Nathaniel Edward Bain, Rebecca Louise Drummond, Charles Alexander Burman, Fiona Nicole Mallory

The proposed neuro-symbolic model combines deep learning and logic-based reasoning to infer diagnoses from multi-modal biomedical data. It addresses ambiguity in clinical evidence interpretation and enhances transparency in complex diagnostic pathways across rare disease and multi-symptom presentation scenarios.

Oscar Thomas Brigham, Emilia Violet Carrington, Hugo Robert McFarlane, Madeleine Ivy Thorpe, Dominic Frederick Weldon

This work develops a CNN-based diagnostic model for pediatric diseases, focusing on underrepresented cases using synthetic oversampling techniques. The method improves detection rates in rare pediatric conditions, mitigating bias and improving fairness in AI-driven screening systems across children’s hospitals globally.

Frederick James Morley, Olivia Beatrice Withers, Leo Harrison Cantrell, Grace Evelyn Pemberton, Rupert John Ashworth

This study leverages graph neural networks to predict potential drug repurposing candidates for COVID-19. By modeling gene-drug relationships across expression and treatment profiles, the model identifies effective agents, accelerating discovery pipelines for emerging pathogens in global health crises.

Isabelle Francesca Redgrave, Julian Peter Channing, Rosalind Kate Fairleigh, Edward Benjamin Musgrave, Alice Miranda Spence

This paper introduces a temporal attention framework to track and predict disease progression using biomedical sensor data. The model emphasizes critical time segments, improving interpretability and accuracy in patient monitoring, especially for chronic disease trajectory forecasting and personalized care planning in clinical environments.

Gareth Lewis Thornbury, Miranda Sophie Layton, Tobias James Huxley, Charlotte Eleanor Weatherby, Felix Matthew Braxton

This study presents a language model tailored to biomedical documents that integrates textual and tabular data. It extracts contextually relevant information from unstructured records, aiding in structured data conversion and enhancing the efficiency of digital health workflows in hospitals and research institutions.

Beatrice Joanna Langford, Elliot Daniel Whittingham, Frances Helena Rosenthal, Oliver Rupert Gainsborough, Harriet Isabelle Fenwick