This paper presents a federated learning architecture that enables collaborative training of cancer prognosis models across hospitals without sharing patient data. It preserves privacy and leverages decentralized histopathological images, enhancing prediction performance while complying with cross-border data governance regulations.

Claudia Frances Bellamy, Patrick Laurence Sandoval, Miriam Olivia Brantley, Felix Dominic Holbrook, Grace Eleanor Caldwell

This study proposes a hybrid CNN-GRU approach to analyze real-time respiratory signals from wearable sensors. The system enhances pulmonary condition assessment by detecting anomalies in breathing patterns, supporting remote monitoring of patients with chronic respiratory illnesses using cost-effective technologies.

Benjamin Louis Ashcroft, Emilia Charlotte Dalrymple, Christopher Maxwell Penn, Florence Helena Rutherford, Sebastian Claude Whitaker

This research develops an interpretable AI framework using Explainable Boosting Machines to classify breast tumors from 3D mammograms. It enhances clinical trust in AI-based diagnoses by revealing decision logic and highlighting key features that influence predictions in real-world screening programs.

Georgia Marianne Clancy, Julian Eric Forsythe, Daphne Louise Ellwood, Rupert Charles McAllister, Imogen Alice Kinsley

This paper introduces a transformer-based time series model for automated cardiovascular risk assessment using continuous ECG recordings. The approach outperforms traditional methods in identifying high-risk patients early, enhancing prevention strategies through precision medicine in cardiac care environments.

Rupert Nathaniel Collingwood, Annabelle Juliet Firth, Gregory Stuart Manning, Harriet Elise Montague, Oliver Francis Quimby

This study builds a reinforcement learning-based decision support system that assists clinicians in managing Type 2 diabetes. By learning from patient histories and outcomes, the model optimizes treatment recommendations, enhances glycemic control, and reduces complications in primary care environments.

Frederick Thomas Langley, Victoria Iris Meadows, Jasper Malcolm Rowntree, Rebecca Eliza Norwood, Samuel Henry Wycliffe

This study proposes a graph neural network model that integrates biomedical knowledge graphs to predict drug-target interactions. The model improves accuracy and interpretability by leveraging molecular relationships and pathway information, aiding drug discovery and repurposing in pharmacological research environments.

Lawrence Benedict Carrington, Phoebe Olivia Yardley, Thomas Gareth Ashford, Amelia June Wakefield, Henry Francis Billings

This paper presents a self-supervised contrastive learning framework for fusing multimodal biomedical images in tumor detection. By learning joint representations across imaging modalities, the model enhances detection accuracy and reduces dependency on labeled datasets, improving diagnostic capabilities in oncology applications.

Felicity Rose Harcourt, Nicholas Hugh Bannister, Eleanor Grace Vickers, Matthew Julian Hollingsworth, Zachary Elliot Frampton