Low-rank tensor decompositions reveal coupled spatiotemporal patterns of CSF tracer transport in the human brain

The study employs intrathecal contrast-enhanced magnetic resonance imaging (MRI) to examine cerebrospinal fluid (CSF) transport in the human brain. This innovative approach, known as glymphatic MRI (gMRI), offers a unique perspective on the interaction between interstitial fluid (ISF) and the glymphatic system. Traditionally, longitudinal data from multiple subjects has been analyzed using univariate comparisons, which disregards the inherent multi-way structure and obscures the dynamic relationships between variables.

To overcome this limitation, the researchers utilized unsupervised non-negative low-rank tensor decompositions to simultaneously analyze the tracer signal in both CSF and brain parenchyma. This method was applied to data collected from 92 patients, comprising 43 individuals diagnosed with idiopathic normal pressure hydrocephalus (iNPH) and 49 reference subjects (REF). The analysis was further stratified by sex to prevent the confounding effects of diagnosis and the observed sex imbalance in the cohorts.

The decomposition process yielded four replicable components, three of which correspond to distinct transport pathways. The first component reveals the distribution of tracer within the supratentorial subarachnoid space and cerebral gray and white matter after a 24-hour period. The second component captures a transient early influx in regions that align with the transport along major cerebral arteries. The third component, which reflects tracer influx in ventricular CSF, also referred to as ventricular reflux.

Notably, the first two components exhibit a positive correlation, establishing a direct quantitative link between the early-stage influx and the 24-hour tracer distribution. Moreover, the expression patterns of these two influx pathways differ significantly between the REF and iNPH cohorts. This finding underscores the utility of unsupervised tensor decompositions as an automated, data-driven approach for stratifying patient groups and deriving quantitative markers of solute transport in the human brain.