Generating virtual populations of cardiovascular shapes for in-silico trials of medical devices

In-silico trials of medical devices is an emerging technology that holds the power to revolutionise the design, selection and deployment of devices used to treat pathologies such as aneurysms, vessel stenosis, valve stenosis, etc. A pre-requisite for conducting systematic, large-scale in-silico trials of medical devices is the development of generative models of anatomy and physiology, representative of real-world patient populations. The focus of this project is on developing novel approaches for conditional generative modelling of cardiovascular anatomy, for in-silico trials of Transcatheter Aortic Valve Implantation (TAVI) devices, used to treat aortic valve stenosis. Specifically, this project will explore generative shape compositional learning as a means to effectively leverage cardiac shapes/structures available in disparate, incomplete, multi-modal imaging data, to generate coherent, anatomically plausible cardiovascular shape populations that are conditioned on relevant patient attributes and representative of real-world patient populations. As such, the project will involve development of: (a) a novel approach to segment aortic valves in contrast-enhanced cardiac computed tomography images; (b) novel generative shape models using shape composition learning; (c) novel conditional generative shape models to facilitate the instantiation of cardiovascular anatomies according to desired attributes in virtual patient populations, for in-silico trials of TAVI devices.