By: Ellis van Etten, Bart van der Worp, Mira Katan

twitter: #GSSW @Ellis_van_Etten

ESO Garmisch Stroke Science Workshop 2021

Session 5: Future Research Models


The last day of the Garmisch Stroke Science Workshop started with a sessions that focused on future research models. The speakers informed the audience about new technical developments and exciting new frontiers in stroke research. This led to a lively discussions about the benefits, but also potential pitfalls of using these new technologies. This session was hosted by Bart van der Worp and Mira Katan.

The first talk and the keynote lecture of this session, was given by Philippe Ryvlin, professor of Neurology at the University of Lausanne. Professor Ryvlin introduced us to the Human Brain Project (HBP) and explained its broad infrastructure and capacity to model and simulate multiscale human brain networks. He elaborated on the steps leading from a classical input, i.e. MRI, to simulation of brain activity and refining the model by applying it to real data. Objectives for stroke research are to better understand the heterogeneity between patients, predict post-stroke recovery and individualize rehabilitation approaches. He demonstrated a framework for federated analyses that allows for standard analyses and machine learning techniques in larger datasets while ensuring privacy regulations. The FERES project (Federating European REgistries for Stroke), which is an EAN-ESO-HBP initiative, will focus on federating European national stroke registries. All in all, many promising possibilities of the HBP to help advance stroke research and improve stroke care.

Moving on, Jeannette Hofmeijer from the University of Twente introduced us to induced pluripotent stem cells for penumbra research. She pointed out that conventional imaging techniques sometimes show surprisingly normal images of damaged brain tissue, e.g. in the case of hypoxia after cardiac arrest. She pointed out that using the human brain-on-a-chip model provides an unique opportunity to translate findings between animal models and patients by creating a network of different types of brain cells. By inducing metabolic conditions that simulate hypoxia, she and her research group can study reversable and irreversible neuronal processes such as synaptic failure and loss of neuronal network activity. Currently, she is using this exciting technique to study potential treatment strategies for ischemic brain damage and we will hear many more interesting insights from her side in the near future.

The third speaker, Peter Kelly from the Mater University Hospital/University College Dublin and our next ESO President, presented his views on future therapeutic targets for secondary prevention. Prof. Kelly started by showing that the residual risk of cardiovascular complications after ischemic stroke or TIA is still substantial despite current secondary prevention strategies. He discussed the optimal blood pressure target according to the current guidelines and emphasized the need for data on safety and generalisability before we can apply these targets to all stroke patients. As for LDL, we are also in need of stroke specific RCT data reflecting the stroke population that is often older and might have underlying small vessel disease. He introduced the concept of adaptive platform trials, such as the STARMAP study, that can answer several research questions simultaneously by performing multiple randomized comparisons, thereby saving precious time, money, and resources.

Lastly, Ivo Jansen from the Academic Medical Centre in Amsterdam gave a presentation about artificial intelligence and the future of imaging. He explained the concept of convolutional neural networks and how this technique cannot only be used to perform classical human tasks, for example spotting an hemorrhage on CT, but also “non-human” tasks that cannot be performed by human assessor. He demonstrated several non-human tasks including vessel segmentation for determining clot characteristics, collateral imaging, and hemorrhage volume quantification. Also pre-hospital triage and treatment planning were mentioned as potential applications. He also pointed out that, although these techniques are very promising, we should keep in mind the risk of bias and the need of enough, good quality training data.