MULTIMODAL-HEALTH: Generalizable AI Models for Pediatric and Adult Clinical Applications

Healthcare data is inherently multimodal, combining images, laboratory results, clinical notes and physiological measurements. However, most artificial intelligence based models remain siloed in single modalities and perform poorly when deployed across hospitals, populations or age ranges. This project develops foundation-model-style multimodal architectures designed to be generalizable across paediatric and adult cohorts and clinical tasks. To ensure the trustworthy deployment of these models in clinical settings, this project focuses on developing computational tools that are easy to interpret, fair and bias-mitigated. By unifying signals and medical structures, this research aims to build the next generation of clinically relevant AI models that support decision-making in diagnosis, monitoring and treatment across a wide range of conditions.