Three dimensional simulations of embolic stroke: clinical comparisons and an equation for sizing emboli from imaging
There is a need to develop Monte Carlo simulations of stroke to run in-silico trials to replace animal models, explore clinical scenarios to develop hypotheses for clinical studies and for interpreting clinical monitoring. We perform three-dimensional (3D) stroke simulations, carrying out in-silico trials to relate lesion volume to embolus diameter and calculate probabilistic lesion overlap maps, building on our previous Monte Carlo method. Simulated emboli were released into a 3D in silico vasculature, supplying gray and white matter brain volumes, to generate individual lesion estimates and probabilistic lesion overlap maps. Computer generated lesions were assessed by clinicians and compared with real world radiological images. Simulations of large single emboli reproduced similar middle cerebral artery (MCA), posterior cerebral artery (PCA) and anterior cerebral artery (ACA) lesions to those observed clinically. A proof-of-concept in-silico trial led to a conjecture relating estimated infarct volume as a percentage of total brain volume to relative embolus diameter: $\mathrm{relative diameter} = [\% \mathrm{infarct volume} / a]^{1/b}$, where $a= 104.2 \pm 0.98$, $b=3.380 \pm 0.030$. Probabilistic lesion overlap maps were created, confirming the MCA territory as the most probable resting place of emboli in the computational vasculature, followed by the PCA then ACA. The article shows proof of concept for developing a 3D stroke model from an automatically constructed vasculature.