This research proposes an interpretable graph neural network model for identifying disease subtypes by integrating clinical records and molecular signatures. The system clusters patients by phenotype-genotype relations, enabling more personalized treatments and advancing the understanding of heterogeneous diseases in biomedical informatics.

George Matthew Covington, Olivia Renee Blackwell, Noah Christian Whitmore, Lydia Rose McKenna, Benjamin Marcus Falkner

This paper introduces a deep attention-based network for detecting hypoglycemic episodes using continuous glucose monitoring (CGM) data. The method identifies subtle glucose variation patterns to enable early intervention, significantly reducing emergency cases in type 1 diabetic patients under real-time monitoring conditions.

Chloe Elizabeth Jennings, Ethan Gregory Wilton, Rebecca Anne Carmichael, Oliver Dominic Harris, Eleanor Madeline Thorne

We propose a hybrid AI framework integrating radiology and genomic data to improve cancer prognosis accuracy. The model aligns imaging features with gene expression vectors to capture mutation-specific progression pathways, offering clinicians personalized outcome prediction and treatment planning support across multiple tumor types.

Henry Maxwell Davenport, Isabelle Nicole Monroe, Edward James Halbrook, Victoria Kate Sanderson, Lucas Alexander Flint

This study presents a cross-domain few-shot learning approach to diagnose rare diseases using limited annotated examples. Meta-embedding fusion across clinical modalities enables robust generalization, offering accurate predictions where data scarcity traditionally hampers diagnosis and treatment of rare and underrepresented patient conditions.

Charlotte Iris Wainwright, Isaac Henry Broughton, Phoebe Louise Marston, Julian Patrick Kingsley, Amelia Florence Denton

This work introduces a lightweight temporal convolutional neural network architecture for real-time classification of EEG signals to predict epileptic seizures. The model maintains high accuracy on constrained medical devices, enabling proactive patient care in resource-limited neurological monitoring environments with continuous data streams.

Jonathan Elias Redmond, Megan Charlotte Fairchild, Zachary Isaac Morton, Alice Penelope Brookshire, Thomas Oliver Wakefield

This research presents a federated learning framework to predict cardiovascular risks using distributed electronic health record data. The approach ensures patient privacy while improving model generalization by aggregating insights from diverse clinical environments without centralizing sensitive medical information.

Daniel Christopher Whitmore, Natalie Louise Bradford, Owen Everett Langston, Georgia Isabelle Meadows, Elijah Preston Carrington

This paper develops a transformer-based architecture for detecting Alzheimer’s biomarkers by fusing PET imaging and genomic sequences. The model identifies early-onset neurodegeneration signatures, facilitating timely interventions and personalized therapeutic strategies to delay or mitigate disease progression in high-risk patients.

Gabriel Louis Pennington, Emma Faith Hollander, Jacob Nathaniel Locke, Victoria Rose Lambert, Samuel Elias Fenwick