AI Helps Show How the Brain’s Fluids Flow

New AI-powered techniques could help researchers better understand Alzheimer’s disease, stroke, and other neurological conditions

The human brain has its own waste-clearing system, constantly moving fluid through and around brain tissue to remove harmful proteins and maintain healthy function. Scientists increasingly believe disruptions in this process are associated with diseases such as Alzheimer’s, stroke, hypertension, and traumatic brain injury. Yet measuring how these fluids move throughout the brain has remained one of neuroscience’s most difficult challenges.

At the University of Rochester, Professor Douglas Kelley and his team are using artificial intelligence and advanced MRI imaging to reveal these hidden flows in unprecedented detail. With support from Empire AI, the researchers are developing new methods that could one day help scientists detect disease earlier, improve drug delivery to the brain, and better understand how neurological disorders develop.

The Challenge

Researchers know that the brain’s glymphatic system uses cerebrospinal fluid to help clear away waste products, including proteins associated with Alzheimer’s disease. However, directly measuring fluid transport throughout an entire living brain is extraordinarily difficult.

While MRI scans provide detailed three-dimensional images of the brain, they cannot directly measure the extremely slow fluid movements that occur within brain tissue. Traditional experimental techniques can observe fluid flow in small regions but cannot capture a complete picture of how the system works across the entire brain.

Without a way to measure these flows, scientists have lacked critical information about how the brain removes waste, delivers therapies, and responds to disease or injury.

The Approach

Kelley’s team combines MRI imaging with a technique known as artificial intelligence velocimetry, using physics-informed AI models to estimate how fluids move through and around the brain. By analyzing MRI data, the AI can determine fluid flow velocities and tissue permeability—measurements that cannot currently be obtained through conventional imaging methods.

The result is a new way to visualize and quantify the brain’s fluid circulation system, giving researchers a much more complete understanding of how these processes function in both healthy and diseased brains. Researchers can now use physics-informed artificial intelligence to determine fluid flow velocities from magnetic resonance imaging (MRI) data, helping reveal aspects of brain function that were previously inaccessible.

Why Empire AI Matters

Training these sophisticated AI models requires substantial computing power.

Before Empire AI, the team relied on conventional resources, where a single dataset required approximately 54 hours of training time and large studies were often constrained by computing bottlenecks and queue delays. Completing the full 14-dataset study would have required more than 31 days of continuous computing time.

With Empire AI, researchers can train large-scale physics-informed neural networks more efficiently, run multiple datasets in parallel, and significantly reduce waiting times for computing resources. What was previously impractical is now achievable, allowing the team to explore larger questions and accelerate discovery.

What Researchers Are Learning

The work has already uncovered new insights into how the brain’s fluid transport system operates:

• The glymphatic system appears to clear particles such as amyloid beta proteins through connected pathways of two distinct types, one of which operates much faster than the other.
• One pathway moves fluid relatively quickly through open spaces around the brain and major blood vessels.
• A second pathway carries fluid through deeper brain tissue at speeds roughly 50 times slower.
• These findings provide one of the clearest pictures yet of whole-brain fluid circulation.