Medical imaging based in silico head model for ischaemic stroke simulation

Stroke is one of the most common causes of death and a leading factor of disability in adults worldwide. It occurs when the blood supply to part of the brain is significantly reduced, potentially leading to the formation of brain oedema. Owing to the rigid nature of the skull, brain expansion results in the shifting of tissue structure, often captured by measurement of the midline shift (MLS). Clinically, MLS has been used in practice as an indication of stroke severity, potential tissue damage and as a way to assess whether decompressive surgery should be performed. However, a growing body of research points towards limitations in such predictive ability. Inspired by the recent progress made in traumatic brain injury simulations, in silico experiments appear as the ideal candidate to elucidate stroke consequences on brain tissues, e.g., morphological changes, in particular in the overarching context of computer model assisted clinical decision making support. To this end, two biologically-informed finite element head models, human and rat, were constructed to support such analysis. The main components of the models include magnetic resonance imaging-derived grey matter, white matter, cerebrospinal fluid and skull, while the human head model also includes the vasculature, additional cerebral components and axonal tractography. Constitutive models representing the mechanical behaviour of each component account in particular for the behaviour of brain tissues during the swelling process accompanying oedema development. The rat model was leveraged for the calibration of the swelling parameters, in turn used for the simulation of human stroke. Human oedema development as a result of stroke was simulated at three frequent locations: basal ganglia, fronto-opercular/anterior insula and temporo-parietal. All three cases exhibit a quadratic MLS evolution with time with the basal ganglia and temporo-parietal showing the largest and smallest values, respectively, at any given time. A proposed injury criterion for axonal tract damage was shown to be larger in the temporo-parietal case. Taken together, these results point towards i) the importance of considering stroke location when using the MLS as an indication of stroke severity, and ii) the potential lack of correlation between MLS value and tissue damage. Ultimately, we propose an in silico methodology that may hold promise in predicting stroke evolution based on an estimate of MLS and stroke location at a given time.