Simulation-based optimisation for the generative design of a shunt.

点击翻译按钮获取中文摘要

This project falls within the EPSRC research area of healthcare technologies while also including topics pertinent to mathematical and engineering sciences. The project will be under the supervision of Prof Antoine Jerusalem (engineering) and Prof Sarah Waters (mathematics) and will work with optimisation tools provided by Prof Jose-Maria Peña through Lurtis Ltd, with clinical guidance provided by Mr Jay Jayamohan. Hydrocephalus is a serious medical condition affecting one in every thousand births. Though arising from various causes, hydrocephalus describes an excess of fluid within the skull, causing a build-up of inter-cranial pressure. If left untreated, the condition will worsen, causing headaches, balance problems, and potentially proving fatal within years. The most common treatment is to implant a permanent drainage shunt into the brain to remove excess fluid to the stomach, where it can be safely cleared. However, this treatment has the risk of vascular brain tissues such as the Choroid Plexus (CP) being dragged into the shunt during drainage, causing both shunt blockages and bleeds in the brain, and additionally preventing the shunt from being easily replaced. A new shunt design is needed to avoid this. The intention of the DPhil is, therefore to create a new, patentable (potentially personalisable) design framework for hydrocephalus shunts, suitable for clinical use. The project will first develop a computational fluid model for the shunt-CP system within the brain, before using generative design principles to optimise the configuration for a given CP idealised geometry. Generative design is a novel approach which combines traditional engineering with advanced artificial intelligence methods (mostly machine learning and heuristic optimisation methods) to produce new tentative designs and refine the ones proposed by the user. Here a heuristic optimisation algorithm developed in partnership with Lurtis is used (based on hybrid self-adaptive optimisation techniques that combine population-based and local search strategies). For this optimisation framework, this DPhil may also explore different surrogate-based approaches that would complement the optimisation mechanism for this particular computational model problem. The model is posed as a fluid-structure interaction problem where outflow through the shunt creates fluid stresses, which causes deflection of the CP tissue. With a sufficiently generalisable mesh, various material parameters can be varied to investigate reduction in this deflection, hence minimising the likelihood of the tissue being entangled in the shunt holes. The optimisation code acts as a black-box wrapper, which can be coupled to the shunt model to investigate parameters of interest and personalised for a given morphology. Work during the short project provided a proof of concept with a simplified 2D model. This project developed an investigative model which successfully simulated the deformation of CP in the hydrocephalus scenario. The optimisation algorithm was used successfully to propose new hole sizes and positions, and suggested that larger diametrically opposite holes would minimise tissue deflection. There is also potential within the scope of a DPhil to create a software tool which could apply these methods to general medical components of a similar remit. We may want to explore and justify our models by comparing simulations to physical experiments run with the engineering department. If the project progresses sufficiently, we may also want to establish contact with medical companies to discuss patent and manufacturing opportunities.