Overview
- A team led by Douglas Kelley published a Science Advances paper describing a physics-informed artificial intelligence technique that extracts fluid velocities and tissue permeability from MRI and dye-spreading videos.
- The study identifies two principal glymphatic flow regimes: faster surface flows of a few microns per second and much slower deep-tissue flows roughly 50 times slower.
- Researchers trained neural networks on time-series dye imaging and embedded physical fluid equations so the model can infer slow, brain-wide flows that conventional MRI and microscopy struggle to measure.
- The team is assembling baseline measurements in mice and plans controlled comparisons across age and disease models while treating human clinical use as a next-stage goal that requires further validation.
- The work was supported by NIH programs and collaborators at the University of Rochester, Brown, and the University of Copenhagen say the approach could eventually help screen for poor brain fluid circulation in conditions such as Alzheimer’s or assess circulation after concussion.