This work develops a federated transfer learning framework for diabetic retinopathy detection using hospital imaging data. The system maintains data privacy while achieving high diagnostic accuracy through collaborative model training across institutions with diverse patient demographics and imaging setups.

Julian Robert Fischer, Emily Kate Darlington, Frederick Isaac Monroe, Chloe Helena Granger, Tobias Leonard Wright

This study introduces an attention-based recurrent model to analyze EEG sequences for early detection of neurological disorders. By leveraging temporal patterns, it distinguishes between progressive and acute conditions, enhancing clinical decision-making in neurology through explainable, high-resolution time-series interpretation.

Lucas Edward Montgomery, Sophia Grace Eberhardt, Noah Sebastian Clarke, Annabelle Isla Reeves, Liam Thomas Hartley

This paper presents a deep residual learning approach with multi-scale image fusion for precise segmentation of lung infections in CT scans. The technique improves delineation accuracy in COVID-19 and pneumonia cases, enabling faster triage and better radiologist support.

Charlotte Amelie Brooks, Theodore James Whitman, Iris Mae Donaldson, Hugo Zachary Preston, Daisy Eleanor Flynn

We propose a graph-based deep learning model that forecasts hospital resource demands during health crises by modeling geographic and temporal dynamics. The model helps healthcare administrators proactively manage ICU beds, ventilators, and staff allocations across interconnected regional facilities.

Samuel Tobias Bright, Amelia Florence Garner, Henry Jacob Ellington, Matilda Rose O’Connell, George William Bexley

This work proposes a reinforcement learning framework for recommending targeted therapies by aligning molecular profiles of cancer patients. The system adapts over time based on treatment feedback, optimizing therapeutic outcomes while minimizing side effects in precision oncology environments.

Florence Clara Wilkins, Jonathan Rupert Cole, Eliza Hope Redford, Sebastian Oliver Maddox, Violet Theresa Blanchard

This research introduces a hybrid attention model that fuses structured electronic health records and unstructured clinical notes for ICU mortality prediction. The model enhances interpretability and decision-making, providing clinicians timely insights into high-risk patient profiles under critical care.

Margaret Louise Donnelly, Edward Simon Carver, Isabelle Rose Atkinson, Thomas Gabriel Langford, Eleanor Maeve Sinclair

We present a novel vision transformer architecture augmented with temporal contrastive learning for analyzing handheld ultrasound data. The framework enables rapid and reliable cardiac anomaly detection in low-resource settings without compromising diagnostic precision or patient safety.

Beatrice Elinor Chapman, Charles Benedict Rowe, Georgia Lily Hammond, Maxwell Owen Ridley, Arabella Grace Nichols