This study introduces a novel ensemble deep learning framework combining CNN, LSTM, and attention networks for multimodal clinical data analysis. The model improves diagnostic accuracy by integrating imaging and textual data for informed clinical decision-making in complex biomedical environments.

William George Franklin, Chiara Elisabetta Moretti, Nathan Oliver Sullivan, Helena Ingrid Nielsen, Lukas Emil Baumgartner

This paper proposes a reinforcement learning-based adaptive system that tailors rehabilitation programs for orthopedic patients. The system dynamically adjusts exercise difficulty based on real-time biomechanical data, improving recovery efficiency while reducing risk of reinjury in physical therapy contexts.

Benjamin Lucas Fordham, Clara Isabelle Møller, Vincent Hugo Johansson, Teresa Camille Dufour, George Patrick Cummings

This research introduces a temporal-spatial attention-based deep fusion model for biosignal integration in emotion recognition tasks. The framework effectively extracts nuanced emotional states by learning relationships between EEG channel time dependencies and spatial interrelations, improving human-computer interaction systems in healthcare.

Jonathan David Keats, Fiona Annabel Sørensen, Oscar Maximilian Langley, Sylvie Juliette Lefevre, Tomas Rafael Heikkinen

We propose a graph neural network architecture to predict drug-protein interactions in metabolic diseases. The model utilizes molecular graph structures and biological pathways to improve prediction accuracy and aid drug repurposing in complex conditions such as diabetes and metabolic syndrome.

Frederik Nikolai Olsen, Isabella Grace Petersen, Leo Sebastian Müller, Ella Henriette Dubois, Marco Giacomo Benedetti

This paper presents a semi-supervised deep learning framework for detecting chronic respiratory disorders using cough acoustics. The model combines autoencoders and noise-robust classification layers, enabling scalable diagnosis from mobile devices and reducing the dependency on labeled clinical data for training.

Hugo Adrian Mitchell, Sophie Victoria Nielsen, André Lucas Delacroix, Jakob Emil Rask, Matilda Claire Hargreaves

This study introduces Bayesian deep learning models that quantify prediction uncertainty in prognosis tasks using longitudinal electronic health records. The models improve trust and safety in clinical decision-making by incorporating confidence intervals and robust posterior inference in chronic disease progression monitoring.

Dominic Arthur Bradley, Sofia Angelica Thomsen, Lucas Manuel Riedel, Harriet Eloise Beaumont, Julian Matteo Pires

This paper proposes a spatio-temporal transformer architecture for integrating multimodal brain imaging data. It enables early detection of neurodegenerative diseases by capturing progression patterns across structural and functional scans, improving screening performance in conditions such as Alzheimer's and Parkinson's disease.

Edward Sebastian Clarke, Annika Sophie Müller, Matteo Lorenzo Greco, Isabella Helene Larsson, Oliver Charles Redford