Graph Signal Processing Models for Neurovascular Decoupling.

Description du projet

Modèles de graphes de vaisseaux cérébraux pour la compréhension neurovasculaire

Le signal BOLD (Blood Oxygenation Level Dependent) est une technique IRMf clé pour mesurer l’activité cérébrale basée sur le couplage neurovasculaire (CN). Cependant, des facteurs tels que les changements physiologiques et le découplage CN compliquent l’interprétation du signal. Comme il n’existe pas de méthode pour séparer l’activité neuronale des réponses vasculaires, les analyses BOLD négligent souvent le rôle des vaisseaux sanguins. Soutenu par le programme Actions Marie Skłodowska-Curie, le projet GRASPING NERVE vise à développer un modèle graphique des vaisseaux cérébraux pour évaluer la connectivité cérébrovasculaire et intégrer les fluctuations du signal BOLD. Cette approche non invasive améliorera la compréhension du CN et facilitera les études sur son découplage pathologique. Le projet validera le modèle en utilisant des champs d’IRM élevés et ultra-élevés, en comparant les techniques BOLD et non-BOLD.

Objectif

Blood Oxygenation Level Dependent (BOLD) signal is a widespread functional Magnetic Resonance Imaging (fMRI) technique to non-invasively study brain activity, and it relies on the mechanism of Neurovascular Coupling (NC), i.e. changes in cerebral blood flow driven by neuronal activity. However, there are various confounding factors for NC, such as homoeostatic physiological changes, or NC uncoupling driven by certain pathologies. Currently, there is no method to disentangle the information associated with neuronal activity from the vascular response. This is an important issue in both physiological imaging and brain activity investigation, as the two signal sources act as competing confounding factors. Furthermore, although vessels are the main blood distribution system, they are mostly ignored when taking into account BOLD analyses.
In this project, I will first create a graph model of cerebral vessels to assess “cerebrovascular connectivity”. I will then use graph signal processing, a novel signal processing technique based on graphs, to embed BOLD signal fluctuations into the vascular and the more traditional tractography-based graphs to disentangle the propagation of neuronal activity and blood flow in these two pathways. This technique will allow disentangling the components of NC non-invasively, offering new insight on brain activity and neurovascular coupling, and allowing further studies on pathological NC uncoupling, such as pre-surgical imaging for tumour.
I will apply this model at the macro- (whole-brain) and meso- (grey-matter- layers) scale, connecting the properties of NC in functional imaging between them. The application will be validated in high (3 Tesla) and ultra-high (7 Tesla) MRI fields to address its feasibility for both research and clinical application. Finally, BOLD will be compared with non-BOLD functional techniques, such as Arterial Spin Labelling and Vascular Space Occupancy, to shed light into the imaging of NC and brain activity.