This study combines multi-modal fundus image data with ensemble learning techniques to predict diabetic retinopathy progression. The framework includes uncertainty quantification to guide clinical risk stratification and improve treatment scheduling, particularly in high-throughput ophthalmology clinics.

Francesca Helena Bartlett, Patrick Andrew Renshaw, Lucy Catherine Drummond, James Oscar Fuller, Sophia Evelyn MacLeod

This paper proposes a real-time fall detection system for elderly care, combining edge-deployed AI models, ambient sensors, and temporal filtering with graph neural networks. The system enables continuous, non-invasive monitoring, reducing emergency response times and enhancing elderly autonomy.

George Louis Morrison, Matilda Ivy Edwards, Rupert Kenneth Hughes, Lydia Charlotte Jameson, Felix William Barrett

This paper introduces a personalized survivorship care recommendation system for cancer patients. The framework integrates patient history, treatment data, and deep reinforcement learning to suggest optimal post-treatment actions, enhancing quality of life and long-term care efficiency.

Eliza Marianne Thornton, Dominic Hugh Rees, Annabelle Clara Vaughan, Benedict Julian Frost, Olivia Harriet Coombs

A novel federated learning framework is developed to facilitate distributed neuroimaging biomarker discovery for Alzheimer’s diagnosis. The model ensures data privacy while enabling collaborative multi-site learning, maintaining performance without centralizing sensitive patient MRI scans.

Benjamin Arthur Whitmore, Harriet Louise Kemp, Oliver Francis Shepherd, Claudia Isobel Fenton, Zachary Matthew Godfrey

This work proposes a multilingual speech analysis model to detect cognitive decline in elderly populations. The method uses contrastive embedding learning and cross-lingual transfer to generalize across languages while maintaining accuracy in diverse cultural contexts.

Christopher Isaac Rowland, Amelia Florence Doran, Hugo Daniel Spencer, Victoria Isabelle Langley, Frederick Joseph Crayton

This paper presents a multi-objective deep learning framework to predict cardiovascular risks using time-series data from wearable sensors. The approach simultaneously optimizes accuracy and interpretability, allowing clinicians to trace decision pathways while achieving robust risk classification across age groups.

Julian Marcus Rothwell, Emily Francesca Waller, Tobias Nathaniel Grove, Isla Catherine Fielding, Leo Charles Stapleton

This study introduces an automated system for polypharmacy risk detection in geriatric populations using medical ontologies fused with graph embeddings. The model predicts adverse interactions by integrating prescription histories with biomedical relations across interconnected knowledge bases.

Clara Juliet Monroe, Samuel Elliot Hargreaves, Madeleine Ruth Pemberton, Jacob Harrison Bloom, Florence Abigail Whitaker