Project

Deep Flow Imaging on a Federated Imaging Biobank to Derive Hemodynamic Biomarkers for Guiding Cerebrovascular Interventions.

Short Summary

Aneurysms and arteriovenous malformations in the brain can cause hemorrhagic strokes, with devastating consequences for the patient. Timely endovascular interventions are key to prevent rupture and hence cerebral hemorrhage in patients, who exhibit an increased risk. To diagnose and classify patients and to guide treatment planning, imaging is of critical importance. The Deep Flow Imaging project will establish significant practical advances of clinical imaging of the brain by providing scalable training databases for federated neural network-based applications, high-speed image reconstruction, data inference and visualization to derive hemodynamic biomarkers to better inform vascular interventions in the brain.

Goals

This project aims to implement Deep Flow Imaging on a Federated Imaging Biobank to make available hemodynamic parameters to improve rupture risk stratification and guidance of cerebrovascular interventions in patients presenting with IAs and AVMs.

Significance

The research is expected to enable significant practical advances of clinical imaging by providing scalable training databases for federated neural network-based applications, high-speed image reconstruction, data inference and visualization to derive hemodynamic biomarkers to better inform vascular interventions in the brain. The software tools will be deployable on our federated imaging biobank prototypes, which keep patient data safe and secure within hospitals, while allowing algorithms and metainformation to travel across domain boundaries and sites with connections to the LeonardMed and BioMedIT infrastructure. The modular concept along with unified exchange protocols and interfaces ensures scalability towards a wider dissemination within the ETH domain and the clinical research landscape.

Background

The prevalence of intracranial aneurysms (IA) and arteriovenous malformations (AVM) ranges between 0.5 and 6%. While the conditions remain asymptomatic in many patients, several patients bear an increased risk of developing hemorrhagic strokes with severe consequences. To guide treatment planning, imaging is of critical importance. While current clinical imaging focuses primarily on angiographic, morphological information, hemodynamic parameters of the cerebral vasculature networks and their modulation are not considered or are extrapolated using computational models with a number of assumptions.