Image analysis and in-silico simulation of acute ischemic stroke

Treatment of acute ischemic stroke (AIS) aims to restore the blood supply in the occluded artery to salvage the brain tissue at risk and improve clinical outcome of patients. We explored image analysis techniques to study the impact, causes, and constituents of ischemic lesion evolution in the subacute time window after treatment. We found that subacute lesion growth is associated with worse clinical outcome. We identified clinical risk factors of subacute lesion growth and showed that subacute lesion growth cannot be explained by edema evolution alone. We also evaluated the applicability of transfer-learning based Convolutional Neural Networks for segmenting follow-up lesions in posterior circulation AIS patients. We also explored in-silico tools to simulate and explain stroke and its treatment. We first described an in-silico trial (IST) platform that can leverage clinical data to generate virtual representations of patients; simulate stroke, its treatment, and the associated brain injury to estimate clinical outcome at population levels, similar to real-life clinical trials. We presented an automated pipeline to develop high resolution, patient-specific finite element vessel and thrombus models from radiological images that could serve as input to the in-silico treatment models. Finally, we validated the IST pipeline with real-life clinical trial data and simulated two exploratory trials to evaluate device performance in different clot types and compare the performance of two devices. Although in-silico techniques will not replace traditional clinical trials, they may potentially contribute towards establishing the efficacy of newer treatments.