The impact of vasculature on fMRI (and NC interpretation) becomes clear considering external factors that can alter brain perfusion and NC, such as individual variations to the vascular system (e.g. incomplete circle of willis) or pathologies with a vascular component (e.g. Moya Moya, dementia, or tumour growth).
Thus, it is important to disentangle BOLD fluctuations solely related to physiological processes from those related to neuronal activity alone, recovering this distinction in cases of altered NC.
Neuronal activity and physiological responses have different spatio-temporal characteristics. Importantly, they travel through the brain via two different mechanisms at different speeds: while neuronal activity traverses white matter (axonal) tracts (WMT) at a faster speed, physiological responses propagate through the vascular system at a much lower speed.
This project aims at leveraging this information to create a vascular-structural model to embed functional MRI signal and disentangle physiological fluctuations from neuronal activity to achieve improved denoising, thereby improved imaging quality, in both cognitive and physiological research.
in order to do so, I will use ultra-high resolution in-vivo vascular imaging to obtain cerebral meso-vessels connectivity, that together with traditional WMT based structural connectivity will be modelled into graphs. Then, leveraging a novel method in signal processing, i.e. Graph Signal Processing, I will filter functional and perfusion MRI timeseries into its WMT and vascular coupled components.
To better approach clinical use, I will use an individual-oriented imaging strategy, dense mapping, collecting many sessions of a few subjects, with a massive multimodal approach, collecting physiological, functional, perfusion, structural, and anatomical MRI data.
The project has four main areas of impact. The first impact will be increasing scientific knowledge about neurovascular coupling, in both health and pathology. On the larger scale, this could be insightful to develop new forms of functional cerebral imaging, and to improve current BOLD functional imaging. The second impact will be expanding the use of Graph Signal Processing to new types of signal and underlying graphs. This might require adapting the current methods, due to the specificity of the data. The third, and biggest, impact will be improving the quality of MRI data and results in general, but specifically in pre-surgical mapping.
By bringing focus on the impact of the vascular system on fMRI (WP1), functional imaging might benefit due to increased specificity and sensitivity, improving MRI-based research and clinical diagnosis and prognosis. While the methods will be developed for cerebral imaging, they will be translatable to non-brain MRI (e.g. cardiac, liver, renal), equally improving data quality. In the long term, by improving the quality of pre-surgical mapping, tumour removal surgery will benefit from improved pre-surgical planning and improved targeting for removal. The fourth impact will be the collection and manual segmentation of ultra-high resolution vascular imaging, that will benefit further work on improving algorithms to perform automatic segmentation, a currently subobtimal and labour intensive process, thus possibly bringing in-vivo meso-vessels imaging to new clinical applications.
