A team from the University of Rochester, Brown University, and the University of Copenhagen has used physics-informed artificial intelligence to pull something out of MRI data that the scans cannot measure directly: how fast fluid moves through the brain. Writing in Science Advances, the researchers describe feeding neural networks videos of dye spreading through brain tissue over time, and from those images deducing both the velocity of the fluid and how permeable the surrounding tissue is.
The target of the work is the glymphatic system, the network that circulates waterlike fluid around the brain during deep sleep and flushes out metabolic waste linked to diseases including Alzheimer's. That system was first described in 2012 by Maiken Nedergaard, co-director of the University of Rochester Center for Translational Neuromedicine, but its mechanics remain incompletely understood - including the basic question of how quickly the fluid actually moves.
The difficulty is methodological. As Kelley explains, a microscope trained on a small patch of brain tissue yields exquisite detail but only a sliver of the overall process, while MRI delivers a three-dimensional view of the whole brain yet cannot capture flow velocity at speeds this slow. Studying circulation inside a living brain without causing lasting damage has been a persistent obstacle. The AI approach is essentially a workaround: use physics to infer what the imaging itself leaves out.
Why it matters for dementia research is straightforward. If waste clearance during sleep is a factor in Alzheimer's, then measuring how well that clearance works - across a whole brain, not just one window into it - is a prerequisite for understanding when and how it fails.