Meeting Banner
Abstract #2985

GPU imaging analysis for ultra-fast non-Gaussian diffusion mapping

Marco Palombo 1,2 , Dianwen Zhang 3 , Chen Zhu 4 , Julien Valette 1 , Alessandro Gozzi 5 , Angelo Bifone 5 , Andrea Messina 6 , Gianluca Lamanna 7 , and Silvia Capuani 6,8

1 CEA/DSV/I2BM/MIRCen, Fontenay-aux-Roses, France, France, 2 IPCF-UOS Roma, Phys. Dpt., Sapienza University, Rome, Rome, Italy, 3 ITG, Beckman Institute, UIUC, Urbana, Illinois, United States, 4 College of Economics & Management, CAU, Beijing, China, 5 IIT, Center for Neuroscience and Cognitive Systems @ UniTn, Rovereto, Italy, 6 Physics Dpt., Sapienza University, Rome, Italy, 7 INFN, Pisa Section, Pisa, Italy, 8 IPCF-UOS Roma, Phys. Dept., Sapienza University, Rome, Italy

The application of graphics processing units (GPUs) for diffusion-weighted NMR (DW-NMR) images reconstruction by using non-Gaussian diffusion models is presented. The image processing based on non-Gaussian models (such as Kurtosis and stretched exponential) currently are time consuming for any application in real-time diagnostics. Non-Gaussian diffusion imaging processing was implemented on the massively parallel architecture of GPUs, by employing a scalable parallel LM algorithm (GPU-LMFit) optimized for the Nvidia CUDA platform. Our results demonstrate that it is possible to reduce the time for massive image processing from some hours to some seconds, finally enabling automated parametric non-Gaussian DW-NMR analyses in real-time.

This abstract and the presentation materials are available to members only; a login is required.

Join Here