Conquering the Challenge of Simulating Blood Flow Through Adaptive Physics Refinement

Simulating how cells move through the human bloodstream requires capturing events that span enormous biological scales—from millimeter-scale blood vessels down to submicrometer interactions occurring on the surface of individual cells. Resolving these interactions in realistic vascular geometries has historically required computational resources so large that many biologically relevant problems remained out of reach.

This challenge sits at the heart of computational medicine. Blood flow is not simply fluid moving through a pipe. It is a dynamic biological environment where red blood cells deform, circulating tumor cells travel through branching vascular networks, and cellular interactions evolve under constantly changing flow conditions.

Capturing these processes accurately requires sophisticated physics models and immense computing power. Yet even the world’s fastest supercomputers have limits. Highly detailed simulations are often too computationally expensive to run at realistic human scales, while faster simplified approaches can miss the biological interactions researchers are trying to study.

For years, researchers were forced to choose between realism and scale. Adaptive Physics Refinement (APR) was developed to help bridge that gap.

A New Approach: Adaptive Physics Refinement

APR is a computational framework that dynamically adjusts the level of physical detail throughout a simulation.

At the center of APR is a finely resolved “moving window” that follows the cell or region of interest through the vascular system. Inside this window, simulations resolve detailed cell-scale physics, including cellular deformation, adhesive interactions, and fine-scale fluid structure interactions surrounding the tracked cell.

Outside the window, the simulation still captures the larger vascular geometry and overall blood flow dynamics, but without the expensive requirement of explicitly resolving every cell-scale interaction throughout the entire domain.

This distinction is critical. The overall flow environment and vessel geometry continue to influence the tracked cell, but the most computationally intensive biological calculations are concentrated only in the local region where those interactions matter most.

Importantly, APR does not simply refine spatial resolution. It selectively refines the physics itself, enabling computationally intensive biological interactions to be resolved only where they are needed.

This adaptive strategy makes it possible to perform biologically realistic simulations at far larger scales than would otherwise be feasible.

Imagine following a single kayaker traveling through an enormous river network. You may not need to study every ripple across the entire river system at the same level of detail, but you do need extremely precise information immediately around the kayaker as they navigate bends, currents, and obstacles. APR works in much the same way: the moving window travels with the cell of interest, preserving detailed local physics while still accounting for the larger flow environment surrounding it.

Behind the Technology

At the core of APR is a combination of advanced mathematical modeling and leadership-class supercomputing.

A single biological simulation may involve billions of interacting variables governed by fluid dynamics, structural mechanics, and molecular interactions. Traditionally, maintaining this level of detail across an entire vascular system quickly becomes computationally prohibitive.

APR introduces a fundamentally different strategy. Rather than resolving detailed cell-scale interactions everywhere simultaneously, the framework concentrates computationally expensive calculations within the moving window while using coarser representations outside the refined region to capture the broader vascular flow field.

The framework was designed specifically for modern heterogeneous supercomputers, where CPUs and GPUs work together to solve different parts of a problem. Memory-intensive tasks can be assigned to CPUs, while highly parallel cell-scale calculations are accelerated on GPUs. This architecture enables simulations that would otherwise require prohibitively large memory footprints and runtimes.

Recent developments have further expanded APR into a scalable platform capable of running thousands of simultaneous cell transport simulations in parallel. These advances include new multi-window formulations, GPU-accelerated adhesive dynamics calculations, deterministic parallel execution strategies, and communication-efficient methods optimized for exascale computing systems such as Aurora at Argonne National Laboratory.

Together, these advances transform APR from a proof-of-concept modeling strategy into what is effectively a computational microscope capable of studying cellular behavior across tissue-scale vascular environments.

From Blood Flow to Cancer Transport

The earliest demonstrations of APR focused on resolving detailed blood flow and cellular interactions within complex vascular geometries.

Building on that foundation, APR was later extended to model blood flow with physiologically realistic red blood cell concentrations. In Enhancing Adaptive Physics Refinement Simulations Through the Addition of Realistic Red Blood Cell Counts, the framework enabled researchers to investigate how red blood cell interactions influence circulation and vascular behavior under more realistic physiological conditions.

These studies marked an important step toward bridging cellular-scale interactions with larger cardiovascular dynamics.

Tracking Cancer Cells Through the Bloodstream

APR has also been extended beyond cardiovascular applications into cancer research.

Metastasis—the spread of cancer cells throughout the body—is responsible for the vast majority of cancer-related deaths. One of the most critical stages of metastasis occurs when circulating tumor cells travel through the bloodstream and interact with distant vascular sites.

These events are exceptionally difficult to study experimentally and computationally. Circulating tumor cells are extremely rare; their interactions with vessel walls are highly stochastic, and their trajectories depend on both fluid dynamics and microscopic adhesive behavior.

Understanding these processes requires the ability to simulate thousands of possible cellular trajectories simultaneously while preserving submicrometer adhesive dynamics around each tracked cell.

In High Performance Adaptive Physics Refinement to Enable Large-Scale Tracking of Cancer Cell Trajectory, APR enabled large-scale simulations of circulating tumor cell transport through complex vascular geometries. By combining tissue-scale transport with detailed local cellular interactions inside the moving window, the framework allowed researchers to investigate how flow, vascular geometry, and receptor binding collectively influence metastatic transport.

These capabilities were previously computationally impractical using conventional fully explicit simulation approaches.

Why APR Matters

The implications of APR extend far beyond computational science.

Many diseases are governed by interactions occurring across multiple scales simultaneously—from whole-organ blood flow down to individual receptor-ligand binding events. APR provides researchers with tools capable of connecting these scales within a single computational framework.

By enabling simulations at unprecedented scale and detail, APR creates new opportunities to study disease progression, identify risk factors earlier, and evaluate therapies more precisely.

APR is also an important enabling technology for the Center for Computational and Digital Health Innovation’s broader Find, Track, Treat mission.

Looking Ahead

The future applications of APR extend well beyond vascular modeling.

The same adaptive framework could be applied to airflow in the lungs, immune cell interactions, or the behavior of implanted medical devices such as stents and artificial valves. More broadly, APR represents a growing shift in computational medicine toward integrated multiscale models capable of connecting cellular behavior, tissue mechanics, fluid transport, and patient physiology within unified computational systems.

As computational resources continue to evolve, adaptive approaches like APR will become increasingly important—not simply because simulations are becoming larger, but because the biological questions themselves are becoming more interconnected.

Adaptive Physics Refinement is a reminder that scientific breakthroughs do not always come from simply using more computational power. Sometimes, they come from using it more intelligently.