Characterizing the spatial patterns and determinants of cerebrospinal fluid pseudorandom flow in the human brain with low b-value diffusion MRI

Arash Nazeri, Hossein Hosseini, Taher Dehkharghanian, Kevin E. Lindsay, Pamela LaMontagne, Joshua S. Shimony, Tammie L.s. Benzinger, Aristeidis SotirasImaging 2025 · DOI

TL;DR

This paper introduces a novel framework, CSF pseudo-diffusion spatial statistics (CYSS), to map and analyze cerebrospinal fluid (CSF) flow dynamics across the entire brain using low b-value diffusion MRI. They identified distinct CSF "waterways" and found that their flow characteristics are influenced by age, sex, brain structure, and blood flow, with aberrant patterns potentially linked to incidental clinical findings.

What Problem Does This Paper Address?

Cerebrospinal fluid (CSF) plays a crucial role in brain health, acting as a shock absorber and waste removal system. Impairments in CSF flow are linked to various neurological disorders. While techniques like phase-contrast MRI can measure CSF flow, they often have limitations in terms of whole-brain coverage, speed, or sensitivity to slow flow. Low b-value diffusion MRI (low-b dMRI) shows promise for interrogating CSF motion, but understanding the complex spatial organization and underlying drivers of this flow across the entire brain has remained challenging. This paper aims to bridge that gap by developing a comprehensive method to map and analyze CSF flow patterns and their determinants.

Methods

The core of this study is the development of a CSF pseudo-diffusion spatial statistics (CYSS) framework. This approach leverages low-b dMRI to estimate mean pseudo-diffusivity (MΨ), a measure reflecting the magnitude of CSF flow within each voxel.

Here's a breakdown of the key methodological steps:

  • Low-b dMRI Acquisition: The study utilizes low-b dMRI data, which is particularly sensitive to slow CSF motion due to its longer T2 relaxation times.
  • MΨ Map Generation: From the low-b dMRI data, MΨ maps are generated, representing the effective motility of CSF.
  • Voxel-based Analysis Framework (CYSS): This framework involves:
  • Template Generation: Creating a common space for all participants' data using advanced registration techniques.
  • Spatial Smoothing: Applying smoothing to the MΨ maps to improve signal-to-noise ratio and facilitate analysis.
  • Seed-based Correlation Analysis: To understand how CSF flow covaries across different brain regions, researchers selected specific "seed" regions (e.g., Sylvian fissures, premedullary cistern) and analyzed the correlation of MΨ values with other brain voxels. This helps identify regions that tend to have similar flow patterns.
  • Non-negative Matrix Factorization (NMF): To uncover data-driven, unbiased patterns of CSF flow organization, NMF was applied. This technique decomposes the complex covariance structure of MΨ maps into a set of distinct "CSF waterways" (CWWs). The researchers evaluated multiple NMF solutions to find the most reproducible and reliable set of patterns.
  • Replication Analyses: The findings were validated on independent datasets with different low-b dMRI acquisition parameters to ensure robustness.
  • Statistical Modeling: Linear mixed-effects models were used to assess the influence of demographic factors (age, sex), brain anatomy (atrophy, ventricular volumes), and cerebral perfusion on MΨ within the identified CWWs.

Datasets

The study utilized several large neuroimaging datasets:

  • OASIS-3 (Open Access Series of Imaging Studies): This was the primary dataset for discovery and analysis, including two subsets with different low-b dMRI acquisition schemes (multi-b-value and b100). This dataset contains data from cognitively unimpaired individuals and those with varying degrees of cognitive decline.
  • MICA (Microstructure-Informed Connectomics): A dataset of 50 healthy participants used for replication analysis.
  • CDMD (Comprehensive Diffusion MRI): A dataset of 26 healthy participants used for replication analysis, notable for its densely sampled diffusion MRI data.
  • BraFTI (Brain Fluid Translational Imaging Lab): High-resolution 3D-CISS images from this dataset were used to assess fine anatomical structures of CSF spaces.

Key Insights and Ideas

  • Low-b dMRI is a powerful tool for whole-brain CSF flow analysis: The study demonstrates that low-b dMRI, when analyzed with the CYSS framework, can provide detailed, whole-brain maps of CSF effective motility, overcoming limitations of previous methods.
  • CSF flow is organized into distinct "waterways": NMF successfully identified 10 reproducible CSF waterways (CWWs) that represent functionally and anatomically distinct compartments of CSF circulation. This provides a new, data-driven atlas of CSF flow.
  • CSF flow is not uniform; it's regionally specific: The study highlights that different CWWs exhibit distinct flow characteristics (MΨ, anisotropy, variability) and are differentially influenced by biological factors.
  • Demographics and brain anatomy shape CSF flow: Age, sex, brain atrophy, and ventricular size significantly influence MΨ in specific CSF spaces, suggesting a complex interplay between brain health and fluid dynamics.
  • Aberrant CSF flow may be linked to clinical findings: A crucial finding is the association between anomalous CSF flow patterns and the presence of clinically consequential incidental findings on MRI, suggesting that CSF flow analysis could potentially aid in identifying subtle pathologies.

Key Results

  • 10 distinct CSF Waterways (CWWs) identified: Using NMF, the researchers identified 10 reproducible CSF compartments with distinct flow characteristics.
  • High reproducibility and reliability: The NMF solutions showed high mean Adjusted Rand Index (ARI) (0.82 ± 0.018), indicating robust pattern identification.
  • Regional effects of demographics and anatomy:
    • Advancing age was associated with higher MΨ in Sylvian fissure CWWs and the chiasmatic cistern.
  • Women showed higher MΨ in caudal basilar cisterns.
  • Brain atrophy and larger ventricle volumes were associated with lower MΨ in certain CWWs.
  • Increased fourth ventricle volume was linked to higher trans-ventricular CSF flow.
  • Cerebral perfusion influences CSF flow: Higher cerebral blood flow was associated with increased MΨ in peri-arterial CSF spaces.
  • Incidental findings in aberrant flow: Out of 38 participants with outlier CSF flow patterns, five had clinically consequential incidental findings, a significantly higher rate than in other participants (p = 3.04 × 10⁻⁵).

Limitations and Future Work

  • Cardiac Gating: The low-b dMRI data used in this study were not cardiac gated. Future work could leverage gated low-b dMRI to better characterize the dynamic, cardiac-cycle-dependent aspects of CSF flow.
  • Intrinsic CSF Properties: The study acknowledges that intrinsic CSF properties like temperature and viscosity could influence MΨ, though this is likely negligible in healthy populations. Pathological conditions affecting these properties might warrant consideration.
  • Cognitive Impairment: The study's analysis of CSF flow in disease states is limited by the relatively small number of participants with cognitive impairment in the OASIS-3 subsets.
  • Future Directions: The authors suggest that future research should consider factors like brain compliance and viscoelastic properties. They also emphasize the need for large-scale, integrative studies to explore potential CSF flow abnormalities in neurodegenerative disorders and to further investigate the link between aberrant CSF flow and clinical pathologies. The developed CYSS framework and CWW atlas offer a new paradigm for studying CSF flow with potential clinical applications.

Summary updated

Abstract

The circulation of cerebrospinal fluid (CSF) is essential for maintaining brain homeostasis and clearance, and impairments in its flow can lead to various brain disorders. Recent studies have shown that CSF effective motility can be interrogated using low b-value diffusion magnetic resonance imaging (low-b dMRI). Nevertheless, the spatial organization of intracranial CSF flow dynamics remains largely elusive. Here, we developed a whole-brain voxel-based analysis framework, termed CSF pseudo-diffusion spatial statistics ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>C</mml:mi> <mml:mi>Ψ</mml:mi> <mml:mtext>SS</mml:mtext></mml:math> ), to examine CSF mean pseudo-diffusivity <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mo>(</mml:mo> <mml:mi>M</mml:mi> <mml:mi>Ψ</mml:mi> <mml:mo>)</mml:mo></mml:math> , a measure of CSF flow magnitude derived from low-b dMRI. We showed that intracranial CSF <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>M</mml:mi> <mml:mi>Ψ</mml:mi></mml:math> demonstrates characteristic covariance patterns by employing seed-based correlation analysis. Next, we applied non-negative matrix factorization analysis to further elucidate the covariance patterns of CSF <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>M</mml:mi> <mml:mi>Ψ</mml:mi></mml:math> in a hypothesis-free, data-driven way. We identified 10 distinct CSF compartments with high reproducibility and reliability, reflected by a high mean adjusted Rand index with a low standard deviation (0.82 [SD: 0.018]) in split-half analyses of the discovery multimodal aging dataset (n = 187). The identified patterns displayed similar <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>M</mml:mi> <mml:mi>Ψ</mml:mi></mml:math> across three replication datasets. In discovery and replication multimodal aging cohorts (unique n = 264), our study revealed that age, sex, brain atrophy, ventricular anatomy, and cerebral perfusion differentially influence <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>M</mml:mi> <mml:mi>Ψ</mml:mi></mml:math> across these CSF spaces. Notably, of the 35 individuals exhibiting anomalous CSF flow patterns, five displayed clinically consequential incidental findings on multimodal neuroradiological examinations, which were not observed in other participants <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mo>(</mml:mo> <mml:mrow><mml:mi>p</mml:mi> <mml:mo>=</mml:mo> <mml:mn>3</mml:mn> <mml:mo>.</mml:mo> <mml:mn>04</mml:mn> <mml:mo>×</mml:mo> <mml:msup><mml:mrow><mml:mn>10</mml:mn></mml:mrow> <mml:mrow><mml:mo>-</mml:mo> <mml:mn>5</mml:mn></mml:mrow> </mml:msup> </mml:mrow> <mml:mo>)</mml:mo></mml:math> . Our work sets forth a new paradigm to study CSF flow, with potential applications in clinical settings.
BibTeX
@inproceedings{nazeri2025characterizing,
  author = {Arash Nazeri and Hossein Hosseini and Taher Dehkharghanian and Kevin E. Lindsay and Pamela LaMontagne and Joshua S. Shimony and Tammie L.S. Benzinger and Aristeidis Sotiras},
  title = {{Characterizing the spatial patterns and determinants of cerebrospinal fluid pseudorandom flow in the human brain with low b-value diffusion MRI}},
  booktitle = {Imaging},
  year = {2025},
  doi = {10.1162/imag\_a\_00473},
}
Figure