GPU and Distributed Memory Iterative Non-Cartesian MRI Reconstruction

Get Started

Leverage existing iterative reconstruction code with minimal headache.

Get Started Quickly

Use Vagrant to setup a (non-GPU) virtual machine to try implementing iterative reconstructions in C++ in just a few minutes. Check the Documentation for more info.

Leverage MATLAB® code

PowerGrid uses Armadillo to provide high level matrix syntax that is very reminsicent of MATLAB®. Using the tools in PowerGrid, experienced IRT users can port most code in a day or two.

OpenACC Accelerated

PowerGrid uses OpenACC to provide GPU accelerated implementations of the non-Unform FFT and Discrete Fourier Transform. MPI is used to enable using multiple GPUs in one or several machines.

Open Source

PowerGrid is licensed under the University of Illinois/NCSA Open Source License, a permissive, BSD derivative license. For more information, check out tl;dr legal.

Brain MR images with and without field correction

Why Iterative Reconstruction?

MR images are often degraded by physical effects such as off-resonance, susceptibility induced gradients, and patient motion. These effects diminsh the quality of images that clinicians and researchers rely on. Fortunately, the physical models underlying these effects are well understood. Iterative reconstruction uses a physics-based model to correct for unwanted effects, such as field inhomogeneity and patient motion.

MRFIL Group Photo

About MRFIL @ The University of Illinois at Urbana-Champaign

We're the Magnetic Resonance Functional Imaging Lab (MRFIL) at the Beckman Institute at the University of Illinois at Urbana-Champaign. Lead by Prof. Brad Sutton of the Department of Bioengineering, we develop techniques and technologies to enable new applications in Magnetic Resonance Imaging. Much of our research centers around the application of computation to the acqusition and reconstruction of MR images.

Follow @mrfil