The automated WMH segmentation pipeline was developed and applied to over 6000 acute ischemic stroke patients. We created a new, for clinical images viable, deep-learning based brain extraction methodology in clinical magnetic resonance imaging, specifically FLuid Attenuated Inversion Recovery (FLAIR) images. This allowed the estimation of each patient's brain volume, which was utilized in an automated quality control step and, independently, to investigate the effect of brain volume on long-term stroke outcome. In addition, the processing pipeline only requires a rough registration to a template, which increases its utility by reducing the computational cost to under 3.5 minutes per patient (initially 2.5 hours), while increasing its reliability. Outlines of the WMH have been improved using a deep-learning based framework for segmentation, showing good agreement with manual segmentations.
To further improve registration quality with clinical image data, which can further help to reduce the complexity of the WMH segmentation task, we created a new multi-template-based registration framework, where each patient's FLAIR image is first registered to a set of age-specific templates before they are then transferred into a common space. After developing an automated ventricle segmentation algorithm on clinical FLAIR sequences, we are able to assess the quality of registration automatically, based on the overlap of the ventricles of the template and the subject.
Additionally, we explored the spatial disease patterns in the brain, utilizing information of vascular territories. This resulted in the identification of spatial variations, which are affected/modified by different risk factors (such as hypertension).
With a latent variable model, we quantified the often-observed protective mechanism in the brain leading to better post-stroke functional outcome. We extended the idea of "brain reserve", which is widely studied in populations with cognitive decline, by including pre-existing disease burden. The resulting concept characterizes the remaining reserve, "effective reserve", and was estimated in a set of stroke patients where we showed its relation with long-term stroke outcome. In these patients a higher reserve was associated with better outcome.
An extension of the outcome models by including information from the field of brain connectivity, we elucidated the contribution of the structural connectivity backbone, as well as the contributing factors of functional connectivity after stroke. This led to a significant improvement in the 90-day outcome models estimated within the first few days after admission.
All together, this work has already led to 11 scientific journal publications, three of them in collaborative efforts with international researchers. Part of the work was highlighted in the Advances in Motion online portal, and all of it has been communicated to scientific and non-scientific audiences throughout the duration of the action at conferences, as well as small group meetings with former stroke patients.
