79th Annual Meeting of the APS Division of Fluid Dynamics (Nov 22 — 24, 2026)

V4399668: Seeing The Unseen FlowCerebral Blood Flow: 4D Flow MRI and Super-Resolution

Authors
  • Seongju Lee, Hanyang University
  • Seungmin Kang, Hanyang University
  • Kyu-Sun Choi, Hanyang University
  • Je Hyeong Hong, Hanyang University
  • Simon Song, Hanyang University

How can blood flow through cerebral arterial bends and branches be measured and reconstructed in detail? 4D Flow MRI uses MR phase differences to noninvasively measure three-dimensional velocity fields over the cardiac cycle. However, voxel averaging and measurement noise obscure fine-scale flow features. Using in vivo data from a healthy volunteer, we explore physics-informed reconstruction for more detailed visualization of cerebral blood flow. Physics-Informed Normalized Gaussian Splatting with Axes Alignment (PINGS-X) addresses these limitations through super-resolution and denoising. It assigns physical quantities to axis-aligned Gaussians distributed in space and time, combining their normalized contributions into a continuous flow field. Gaussian positions, widths, and physical quantities are optimized to fit low-resolution measurements while reducing residuals of the continuity and Navier-Stokes equations. Splitting, cloning, and merging adapt the representation, which can then be evaluated on a finer grid. From measurements with 0.5 mm isotropic voxels, we reconstruct (2.9 hr/fr, NVIDIA RTX 6000 Ada 48GB) velocity fields on a 0.25 mm isotropic grid and compare streamlines, cross-sectional fields, and velocity profiles in the internal carotid artery siphon. The reconstruction preserves the overall profile shapes and trends while yielding denser, smoother, less noisy velocity fields. Local peaks remain distinct at peak systole and late diastole, showing that super-resolution and denoising can retain both overall flow patterns and local features. Pressure difference estimates from PINGS-X further extend reconstruction to hemodynamic analysis. Magnetic Resonance Velocimetry (MRV) extends the same phase-contrast velocity-encoding principle to engineering flows. Examples inside household appliances and around a car's rear underbody illustrate magnetic resonance measurements as a common tool for 3D flow visualization and analysis in engineered systems.

This work is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. Any reuse must credit the author(s) and provide a link back to this page.