3D phase contrast MRI: Partial volume correction for robust blood flow quantification in small intracranial vessels
Among the various techniques available for monitoring blood flow in intracranial vessels, 3D‐phase contrast MRI (3D‐PCMRI) 3 is the only non‐invasive method able to measure the full 3D velocity field. 3D‐PCMRI was applied extensively to various vascular structures as reported in the review of Markl et al. 10. Unlike 2D‐PCMRI which requires cumbersome manual positioning of the measurement planes, 3D‐PCMRI allows the acquisition of the full volumetric information which is analyzed afterward. This is particularly convenient for investigating blood flow rates in complicated vessel networks such as the highly variable circle of Willis 11. However, volumetric acquisition of the three velocity components requires longer acquisition time compared to 2D‐PCMRI for lower spatial resolution, limiting its clinical use in small intracranial vessels having cross‐sections of only a few voxels.
To overcome this issue, two different approaches have been explored: (1) the use of acceleration techniques such as k‐t BLAST or HYPERFlow 13 improving the spatial resolution while keeping an acceptable acquisition time, (2) the adaptation of the post‐processing methodology taking into account the possible partial volume and Gibbs ringing artifacts in flow rate assessment 15. Mainly tested on 2D‐PCMRI datasets, these post‐processings require the localization of the circulating volume or assume idealized velocity profiles (typically parabolic Poiseuille flow) which are in general erroneous. Previous in vitro 17 and recent in vivo 18 studies have shown that these partial volume effects might lead to blood flow rate overestimation. However, these effects were not considered in most of the recent intracranial flow investigations 20.
In this study, we intend to address this issue by adapting the partial volume correction methodology proposed in Ref. 15 to both 2D‐ and 3D‐PCMRI data. This method allows for correcting blood flow passing through partially flowing voxels without making any assumption on the velocity profile. Instead, the magnitude of PCMRI data providing information on the degree of partial occupancy is used to correct the velocity, accordingly. However, the vessel edge potentially affected by partial volume artifacts is poorly identifiable on 3D‐PCMRI data due to its low contrast and spatial resolution. Therefore, we combine 3D‐PCMRI velocity data with precise vessel geometry provided by time‐of‐flight MRI (3DTOF) or 3D rotational angiography (3DRA). The methodology presented in this article is first tested in vitro on idealized and patient specific models with physiological flow and subsequently applied to intracranial arteries in patients harboring internal carotid artery (ICA) aneurysms.