Despite rigorous testing requirements for medical device approval, around 83,000 patient deaths and injuries have been linked to implanted devices over the past decade. Conventional human trials struggle to recruit sufficiently diverse patients, while animal testing raises ethical concerns and may not accurately represent the target patient population. This project aims to develop and validate advanced computational techniques to synthesise large virtual populations of 3D patient anatomies that are statistically representative of real-world diversity. Unlike current methods limited by patient data availability, our unique generative modelling approach can combine information across multiple datasets to create plausible but fully synthetic anatomies. Importantly, the synthesis process can be controlled to match target patient demographics, enabling more inclusive device evaluation. The project will focus on generating diverse virtual heart models to support simulation-based (in silico) testing of a variety of cardiovascular devices. Success will lead to more affordable, ethical and representative device testing to ensure patient safety while reducing costs and time to market for new medical technologies. Validated synthetic datasets have potential for broader commercialisation to medical companies undertaking regulatory submissions.
