In-silico clinical trials for assessment of intracranial flow diverters

In-silico trials refer to pre-clinical trials performed, entirely or in part, using individualised computer models that simulate some aspect of drug effect, medical device, or clinical intervention. Such virtual trials reduce and optimise animal and clinical trials, and enable exploring a wider range of anatomies and physiologies. In the context of endovascular treatment of intracranial aneurysms, in-silico trials can be used to evaluate the effectiveness of endovascular devices over virtual populations of patients with different aneurysm morphologies and physiologies. However, this requires (i) a virtual endovascular treatment model to evaluate device performance based on a reliable performance indicator, (ii) models that represent intra- and inter-subject variations of a virtual population, and (iii) creation of cost-effective and fully-automatic workflows to enable a large number of simulations at a reasonable computational cost and time.

Flow-diverting stents have been proven safe and effective in the treatment of large wide-necked intracranial aneurysms. The presented thesis aims to provide the ingredient models of a workflow for in-silico trials of flow-diverting stents and to enhance the general knowledge of how the ingredient models can be streamlined and accelerated to allow large-scale trials. This work contributed to the following aspects: 1) To understand the key ingredient models of a virtual treatment workflow for evaluation of the flow-diverter performance. 2) To understand the effect of input uncertainty and variability on the workflow outputs, 3) To develop generative statistical models that describe variability in internal carotid artery flow waveforms, and investigate the effect of uncertainties on quantification of aneurysmal wall shear stress, 4) As part of a metric to evaluate success of flow diversion, to develop and validate a thrombosis model to assess FD-induced clot stability, and 5) To understand how a fully-automatic aneurysm flow modelling workflow can be built and how computationally inexpensive models can reduce the computational costs.