MRI-based computational model generation for cerebral perfusion simulations in health and ischaemic stroke

Abstract

AbstractCerebral perfusion models were found to be promising research tools to predict the impact of acute ischaemic stroke and related treatments on cerebral blood flow (CBF) linked to patients’ functional outcome. To provide insights relevant to clinical trials, perfusion simulations need to become suitable for group-level investigations, but computational studies to date have been limited to a few patient-specific cases. This study set out to overcome issues related to automated parameter inference, that restrict the sample size of perfusion simulations, by integrating neuroimaging data. Seventy-five brain models were generated using measurements from a cohort of 75 healthy elderly individuals to model resting-state CBF distributions. Computational perfusion model geometries were adjusted using healthy reference subjects’ T1-weighted MRI. Haemodynamic model parameters were determined from CBF measurements corresponding to arterial spin labelling perfusion MRI. Thereafter, perfusion simulations were conducted for 150 acute ischaemic stroke cases by simulating an occlusion and cessation of blood flow in the left and right middle cerebral arteries. The anatomical (geometrical) fitness of the brain models was evaluated by comparing the simulated grey and white matter (GM and WM) volumes to measurements in healthy reference subjects. Statistically significant, strong positive correlations were found in both cases (GM: Pearson’s r 0.74, P-value< 0.001; WM: Pearson’s r 0.84, P-value< 0.001). Haemodynamic parameter tuning was verified by comparing total volumetric blood flow rate to the brain in reference subjects and simulations resulting in Pearson’s r 0.89, and P-value< 0.001. In acute ischaemic stroke cases, the simulated infarct volume using a perfusion-based proxy was 197±25 ml. Computational results showed excellent agreement with anatomical and haemodynamic literature data corresponding to T1-weighted, T2-weighted, and phase-contrast MRI measurements both in healthy scenarios and in acute ischaemic stroke cases. Simulation results represented solely worst-case stroke scenarios with large infarcts because compensatory mechanisms, e.g. collaterals, were neglected. The established computational brain model generation framework provides a foundation for population-level cerebral perfusion simulations and for in silico clinical stroke trials which could assist in medical device and drug development.