GENIUS: Grid Enabled Neurosurgical Imaging Using Simulation

点击翻译按钮获取中文摘要

Cardiovascular disease is the cause of a large number of deaths in the developed world; conditions such as arterio-venous malformations and aneurysms can lead to strokes and death. Problems are often related to anomalous cerebral blood flow behaviour in regions of arterial branching within the brain; therefore, control of cerebral blood flow is crucial in the understanding, diagnosis and treatment of cardiovascular disease. Currently such flow details are not well understood, while experimental studies of cerebral blood flow are often impractical due to the difficulties in measuring real time blood flow in humans. X-ray and magnetic resonance imaging angiography (MRA) are non-invasive ways to produce static images of cerebral arterial structure, but have difficulty producing dynamic images due to the time scales required to obtain the images, and the need to inject the patient with tracer fluids.Some studies have revealed relationships between specific flow patterns around walls and cardiovascular diseases such as atherosclerosis. Current imaging methods represent a very important tool for diagnosis of various cardiovascular diseases and for the design of cardiovascular reconstructions and devices to enhance blood flow. The development of techniques which produce more accurate dynamical information on the flow of fluids in the blood within the brain will give the surgeon much greater success rates when treating cerebral blood flow pathologies, for example when re-routing flow and injecting cross-linking epoxy resins to block off channels.Modelling and simulation have a crucial role to play in neurosurgical blood flow treatments, due to the limitations of experimental methods. Simulation offers the clinician the possibility of performing non-invasive virtual experiments in order to plan and study the effects of certain courses of treatment with no danger to the patient. Modelling and simulation offer the prospect of providing clinicians with virtual patient specific analysis and treatments. Achieving these goals is dependent on the availability of computational models of sufficient complexity and power. In this project we will use the brain imaging techniques discussed above to provide input data for such simulations. The computational requirements to perform such simulations are huge in terms of the memory and number of processors required, meaning that we have to distribute them across multiple high performance supercomputers. Using our home grown application for modelling cerebral blood flow, HemeLB, we will perform such simulations on a federated grid of supercomputers, using resources provided by both the UK and the US; we shall take advantage of fast network links to enable efficient communication between these resources. We will also use tools to allow us to steer and visualise the simulations as they are in progress. We will take advantage of the ability to co-reserve time on these distributed machines to allow us to launch simulations as and when we require. Such a capability is essential to engage clinicians with high performance computing, allowing them to perform simulations and use resources at times convenient for them. We believe that this project will provide a prototype brain blood flow modelling environment that will be of considerable value to clinicians; we will continue to work beyond the scope of this project to exploit this technology for use in everyday surgical procedure planning.