Quantification of Perivascular Flow Dynamics of Human Cerebrospinal Fluid Along Major Arteries Using Phase-Contrast MRI

Background Dysfunction in cerebrospinal fluid (CSF) flow dynamics is linked to several neurologic disorders. Although phase-contrast MRI has been used to measure them in the brain, studies in the perivascular space are lacking. Purpose To develop a pipeline of image processing algorithms using contrast-unenhanced 7-T phase-contrast MRI to locate and quantify human brain CSF dynamics along the perivascular space of major arteries and their branches. Materials and Methods In this prospective, single-center study, data were collected from healthy participants, enrolled consecutively, who underwent contrast-unenhanced 7-T phase-contrast MRI at the Houston Methodist Research Institute between February 2020 and July 2023. Data preprocessing included skull removal and contrast enhancement. Vessel detection was achieved by using a chained hysteresis algorithm with two tunable parameters: a high threshold for identifying vessel pixels and a low threshold for image noise cutoff. Perivascular space was detected using a dilation algorithm. CSF velocity was measured in the cerebral aqueduct and three manually segmented perivascular regions: region 1 (internal carotid artery and M1 branch of middle cerebral artery [MCA]), region 2 (M2 segment of MCA), and region 3 (distal cortical branches of MCA or M3 segment). Linear mixed-effects models were calculated between CSF velocity curves in the aqueduct and perivascular spaces to assess their relationship. Two-tailed Mann-Whitney U and F tests were used to assess differences in CSF velocity across two age groups (age 20-39 years and 40-59 years). Results A total of 28 participants (mean age, 40 years ± 11 [SD]; 14 male and 14 female participants) were included. A linear mixed-effects model showed a positive association between aqueduct and perivascular CSF velocity in three regions along the MCA (0.013 cm/sec [ P < .001], 0.024 cm/sec [ P < .001], and 0.010 cm/sec [ P < .001]). Additionally, the height (0.06 vs 0.16 cm/sec; P = .002) and SD (0.02 vs 0.05 cm/sec; P < .001) of the perivascular CSF velocity curve in the distal cortical branches of the MCA differed between the two age groups. Conclusion A semiautomated pipeline of algorithms using phase-contrast MRI was developed to locate and quantify age-dependent human CSF dynamics along the perivascular spaces of major brain arteries. © RSNA, 2025 Supplemental material is available for this article.