Comparing performance of many‐core CPUs and GPUs for static and motion compensated reconstruction of C‐arm CT data

Abstract

Algorithmic innovations in the field of motion compensated image reconstruction may lead to a shift back to CPUs in the future. For analytical 3-D reconstruction, the authors show that the gap between GPUs and CPUs became smaller. It can be performed in less than 20 s (on-the-fly) using a 32-core server.

Pushing the limits for medical image reconstruction on recent standard multicore processors

“…For GPUs in particular, large performance gains compared with CPUs were reported (Mueller et al 2007) or documented by the standardized R abbit CT benchmark (Rohkohl et al 2009a). 1 Available studies with standard CPUs indicate that large servers are required to meet GPU performance (Hofmann et al 2011). In this report we also use the R abbit CT environment, which defines a clinically relevant test case and is supported by industry.…”

Fully 3D list‐mode time‐of‐flight PET image reconstruction on GPUs using CUDA

“…Programmable graphics processing units (GPUs) have emerged as a popular alternative to multi-CPU computer clusters for high performance computation because of potential savings of cost, space, and power. 15 A significant research effort has been devoted to exploiting GPU technology for x-ray computed tomography (CT), 16,17 digital tomosynthesis, 18 radiation treatment planning, 19 dose calculation, 20 and other medical physics applications. 21 Compared with multi-CPU clusters, 22 the faster speed and lower cost of communication between GPU cores makes scaling easier for applications that involve internode communication.…”

DenseGPU-enhanced surface reconstruction from stereo endoscopic images for intraoperative registration

“…need for highly parallelizable problems, its limited memory, or the costly memory access. 24 The main contributions of this publication are as follows. We describe in detail a complete redesign of the correspondence analysis method to work on the GPU.…”