This project aims to develop advanced artificial intelligence (AI) models to accurately predict cardiovascular disease (CVD) risk using multimodal data-electrocardiogram (ECG), genomics, proteomics, and imaging biomarkers-from UK Biobank. CVD is the leading global cause of morbidity and mortality, yet traditional risk models lack precision. Integrating AI with rich biomedical data offers the potential for earlier, personalized intervention.
Research questions include:
How can AI models detect early ECG-based signals predictive of future cardiovascular events?
Which genomic and proteomic features significantly improve risk prediction when combined with clinical profiles?
Can machine learning enhance the interpretation of cardiac imaging (MRI, CT) for CVD detection and prognosis?
How can multimodal data integration via AI provide a comprehensive cardiovascular risk stratification?
Objectives:
Develop deep learning algorithms for ECG-based risk prediction.
Identify key genetic and proteomic biomarkers for enhanced model performance.
Apply ML techniques to imaging-derived cardiac phenotypes.
Build explainable, integrative AI models across all modalities for clinical deployment.
Dissemination and compliance with UK Biobank’s AI policy:
Models will be developed and validated with interpretability in mind, using explainable AI techniques (e.g., SHAP, saliency maps). We will publish all findings in peer-reviewed open-access journals, and make AI model code, trained weights, and relevant scripts available on GitHub or institutional repositories in accordance with UK Biobank’s AI policy. The outputs will be presented at public scientific meetings to maximize transparency and utility.