Three dimensional simulations of embolic stroke: clinical comparisons and an equation for sizing emboli from imaging

The size and location of ischaemic lesions provides vital clues to the origins of a stroke. This study performs three-dimensional (3D) stroke simulations for comparison with brain images to predict the size and location of emboli. Simulated emboli were released into a 3D in silico vasculature, supplying grey and white matter brain volumes, to generate individual 3D lesion estimates and probabilistic lesion overlap maps. Computer generated lesions were compared with real world radiological images obtained from the Anatomical Tracings of Lesions After Stroke (ATLAS) dataset by a panel of three clinically qualified experts. Simulations of large single emboli entering the vasculature reproduced similar middle cerebral artery (MCA), posterior cerebral artery (PCA) and anterior cerebral artery (ACA) lesions to those observed clinically. Estimated infarct volume as a percentage of total brain volume was found to be related to embolus diameter according to the following expression: $\mathrm{diameter}=[\% \mathrm{infarct\, volume} / a]^{1/b}$ where $a= 2.54 \pm 0.062 \mathrm{mm}^{-b}$, $b=3.380 \pm 0.030$ (with diameter in mm). Maximum embolus sizes observed were consistent with those observed clinically. 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. We conclude that 3D stroke simulations are capable of reproducing radiologically observed lesion distributions following embolic stroke. Personalised stroke simulations aimed at rapid diagnosis of the location and source of a stroke, have potential to distinguish cardioembolic sources from other types of stroke, facilitating rapid access to targeted treatment. Future work will focus on matching patient-specific simulation results to real-world brain imaging to inform future predictive diagnostics.