Robust data-driven segmentation of pulsatile cerebral vessels using functional magnetic resonance imaging.

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Cardiovascular Physiology

Background:

• Functional magnetic resonance imaging (fMRI) provides insights into brain function, vascular health, and waste clearance.
• Accurate segmentation of cerebral vessels is crucial for advancing fMRI-based fluid dynamics research.
• Current methods struggle with fMRI distortions, limiting reliable vessel identification in fMRI space.

Purpose of the Study:

• To develop a data-driven, automatic method for segmenting cerebral vessels directly within fMRI data.
• To address the challenge of misregistration and distortions inherent in fMRI data.
• To enable reliable investigation of neurofluid dynamics in major cerebral arteries and the superior sagittal sinus (SSS).

Main Methods:

• Developed an automatic segmentation approach leveraging pulsatile signal patterns of vessels during the cardiac cycle.
• Identified large cerebral arteries and the SSS directly in fMRI space, overcoming spatial distortions.
• Validated the method on a local dataset and tested reproducibility using the Human Connectome Project (HCP) ageing dataset (422 participants).

Main Results:

• The method successfully segmented large cerebral arteries and the SSS by utilizing their unique pulsatile signatures.
• High reproducibility was confirmed in the HCP ageing dataset, with intraclass correlation coefficients > 0.7 for both artery and SSS segmentation volumes.
• The approach demonstrated robust performance across a wide age range (36-90 years) with repeated scans.

Conclusions:

• Automatic segmentation of major cerebral arteries and the SSS is feasible directly within fMRI data.
• This novel method overcomes limitations of existing techniques related to fMRI distortions.
• The demonstrated reproducibility facilitates accurate and reliable fluid dynamics investigations in key cerebral vasculature using fMRI.