The IMPACT Research Group, under the direction of Professor Wen-mei W. Hwu, is well known for development of the IMPACT Compiler, which is now widely used in industry and academic research. Their recent work addresses application GPUs for magnetic imaging.
The MRI toolset is an implementation in CUDA for iterative MR image reconstruction using Graphics Processing Units (GPUs). It is used in the research of medical imaging, especially in the area of image reconstruction for magnetic resonance imaging (MRI). In experiments, this implementation significantly reduces the computation times by two orders of magnitude (compared with the non-GPU implementation) while accurately compensating for field inhomogeneity.
The MRI toolset can also benefit to the image reconstruction of other tomography modality, such as positron emission tomography (PET), single photon emission computed tomography (SPECT), computed tomography (CT). We believe the speedup of image reconstruction will significantly influence the development speed of neuroscience (i.e. functional brain imaging study related with depression, memory) and clinical diagnosis (i.e. detection of cancer, heart disease, brain tumor).
MRI toolset is developed and maintained by Magnetic Resonance Functional Imaging Lab at UIUC.
Yue Zhuo, Xiao-Long Wu, Justin P. Haldar, Wen-mei Hwu, Zhi-Pei Liang, Bradley P. Sutton, Multi-GPU Implementation for Iterative MR Image Reconstruction with Field Correction. Proceedings of International Society for Magnetic Resonance in Medicine (ISMRM) 2010. [PDF]
Yue Zhuo, Xiao-Long Wu, Justin P. Haldar, Wen-mei Hwu, Zhi-Pei Liang, Bradley P. Sutton, Accelerating Iterative Field-Compensated MR Image Reconstruction on GPUs. 7th IEEE International Symposium on Biomedical Imaging: From Nano to Macro, ISBI 2010 – Proceedings, Article number 5490112, Pages 820-823 [PDF] [DOI: 10.1109/ISBI.2010.5490112]
We propose a fast implementation for iterative MR image reconstruction using Graphics Processing Units (GPU). In MRI, iterative reconstruction with conjugate gradient algorithms allows for accurate modeling the physics of the imaging system. Specifically, methods have been reported to compensate for the magnetic field inhomogeneity induced by the susceptibility differences near the air/tissue interface in human brain (such as orbitofrontal cortex). Our group has previously presented an algorithm for field inhomogeneity compensation using magnetic field map and its gradients. However, classical iterative reconstruction algorithms are computationally costly, and thus significantly increase the computation time. To remedy this problem, one can utilize the fact that these iterative MR image reconstruction algorithms are highly parallelizable. Therefore, parallel computational hardware, such as GPU, can dramatically improve their performance. In this work, we present an implementation of our field inhomogeneity compensation technique using NVIDA CUDA (Compute Unified Device Architecture)-enabled GPU. We show that the proposed implementation significantly reduces the computation times around two orders of magnitude (compared with non-GPU implementation) while accurately compensating for field inhomogeneity.