The invention is a cost efficient alternative to large and expensive computer clusters.

About

Background: Researchers at Stanford University have developed an accelerated and cost efficient system for tomographic image reconstruction. This invention is a method for projecting lines an order-of-magnitude faster using a high-level compute language that runs on graphics processing units (GPU). In this framework, the lines are projected using “list-mode” (i.e. process individual lines with arbitrary 3D endpoints and shift-varying kernels). This approach can greatly accelerate reconstruction algorithms for several medical imaging modalities, including positron emission tomography (PET), single-photon emission computed tomography (SPECT), and x-ray computed tomography (CT). It is most useful for schemes that require "list-mode" processing, such as time-of-flight, dynamic, or high-resolution PET. The invention is a cost efficient alternative to large and expensive computer clusters. In addition, because it uses “CUDA”, a high-level compute library for GPUs, the approach is highly maintainable and easy to integrate in complex imaging systems. Stage of Research We are currently investigating the GPU- Compute Unified Device Architecture (CUDA) platform and the latest hardware architecture from NVIDIA to further explore the capabilities of the GPUs in iterative list-mode image reconstruction for time-of-flight PET. Preliminary investigations show that the GPU-CUDA approach has the potential of further speedup using resources that are only available to CUDA, and is more flexible, scalable, and easy to manage than alternative approaches. Ongoing Research: Full time graduate student hired to further research GPU acceleration of line projections with CUDA. An implementation based on GPU-CUDA platform was completed, and performance tuning is underway. We are planning to test the new NVIDIA Fermi hardware infrastructure, which has promising speedup potential. Shift-varying projection kernels and more accurate physical models are being embedded into the reconstruction process to further enhance image quality and increase accuracy while maintaining high computational efficiency of the image reconstruction process.  Applications: Medical Imaging: Iterative image reconstruction for PET, SPECT, and CT. Particularly advantageous for "list-mode" reconstruction schemes, such as time-of-flight, dynamic and high-resolution PET, with shift-varying system models for enhanced image quality and accuracy without the need for CPU clusters. Advantages: Fast, accurate, and cost-efficient: Method uses a graphics processing unit (GPU) to accelerate line projections. CUDA, a high-level computation library for GPUs, is used with GPU to perform shift-varying projections in list-mode in order to greatly accelerate the image reconstruction process. Simple, easy to maintain: The method is an improvement over a previously-developed list-mode projection method because it uses CUDA, a simpler, easier-to-maintain language, rather than OpenGL, for performing the calculations on the GPU. Also, unlike OpenGL, CUDA provides acces  

Register for free for full unlimited access to all innovation profiles on LEO

  • Discover articles from some of the world’s brightest minds, or share your thoughts and add one yourself
  • Connect with like-minded individuals and forge valuable relationships and collaboration partners
  • Innovate together, promote your expertise, or showcase your innovations