Menu:

GPU-Accelerated Toolkit for Image Reconstruction and Fusion

Deformable image registration is an enabling technology used for specialized applications in treatment planning, intervention, and treatment verification. Ideally, deformable registration would replace rigid registration for routine clinical use and improve the geometric precision of a variety of medical procedures. However, modern implementations are time consuming, give unpredictable results, and are difficult to evaluate. Therefore, despite the presence of known organ deformation, rigid registration is still preferred for many image fusion applications due to its simplicity, robustness, and ease of control.

This research effort aims to remedy the two major limitations of modern deformable registration methods, namely, speed and robustness, through the use of multi-core processors, including GPUs. While the speedup varies by application, results indicate that image reconstruction and registration algorithms designed for a GPU typically run between 20 to 80 times faster than the corresponding CPU implementations, and with near-identical result quality. With this level of speedup, algorithms that previously took minutes (or hours) to complete can now be run in seconds (or minutes). Data-parallel versions of four widely-used deformable registration algorithms are being developed using the Compute Unified Device Architecture (CUDA), a programming abstraction for general-purpose computing on GPUs, introduced by NVidia. These algorithms are based on: (1) demons optical flow, (2) fluid registration, (3) thin-plate splines, and (4) B-splines.

Source Code for the Plastimatch Registration Toolkit

The complete source code for the GPU kernels for FDK-based CT reconstruction, demons registration, and B-spline registration can be downloaded under an open-source license from the Plastimatch project web site.

Publications

Journal Papers

  • J. Shackleford, N. Kandasamy, and G. Sharp, "On Developing B-spline Registration Algorithms for Multi-Core Processors," Physics in Medicine and Biology, vol. 55, no. 21, pp. 6329-52, December 2010. Featured article. Download from IOP Science.
  • G. C. Sharp, N. Kandasamy, H. Singh, and M. Folkert, "GPU-Based Streaming Architectures for Fast Cone-beam CT Image Reconstruction and Demons Deformable Registration," Phys. Med. Bio., vol. 52, no. 19, pp. 5771-5783, September 2007. Download paper.

Book Chapters

  • J. Shackleford, N. Kandasamy, and G. Sharp, "Deformable Volumetric Registration using B-splines," GPU Computing Gems 4, W-M. Hwu (Editor), Elsevier, Dec. 2010.

Conferences

  • J. Shackleford, Q. Yang, A. Louren, N. Shusharina, N. Kandasamy, and G. Sharp, "Analytic Regularization of Uniform Cubic B-spline Deformation Fields," Proc. 15th Int'l Conf. Medical Image Computing & Computer Assisted Intervention (MICCAI), 2012. Acceptance ratio: 32%.
  • G. Sharp, R. Li, J. Wolfgang, G. Chen, M. Peroni, M. Spadea, S. Mori, J. Zhang, J. Shackleford, and N. Kandasamy, "Plastimatch: An Open Source Software Suite for Radiotherapy Image Processing," Proc. Int'l Conf. Computers in Radiation Therapy (ICCR), May 2010.

Workshops

  • J. Shackleford, N. Kandasamy, G. Sharp, "Accelerating MI-based B-spline Registration using CUDA Enabled GPUs," Proc. MICCAI 2012 Data- and Compute-Intensive Clinical and Translational Imaging Applications (DICTA-MICCAI), 2012.

  • J. Shackleford et al., "Plastimatch 1.6: Current Capabilities and Future Directions," Proc. MICCAI 2012 Image-Guidance and Multimodal Dose Planning in Radiation Therapy Workshop, 2012.

Short Papers and Posters 

  • J. Shackleford, G. Sharp, and N. Kandasamy, "Accelerating the Feldkamp, Davis, and Kress Back-Projection Algorithm Using GPUs," Int'l Conf. Computers in Radiation Therapy (ICCR), 2009.

  • G. Sharp, R. Li, J. Wolfgang, G. Chen, M. Peroni, M. Spadea, S. Mori, J. Zhang, J. Shackleford, and N. Kandasamy, "Plastimatch – An Open Source Software Suite for Radiotherapy Image Processing," Int'l Conf. Computers in Radiation Therapy (ICCR), 2009.

  • G. Sharp, Z. Wu, and N. Kandasamy, "A Data Structure for B-Spline Registration," Annual Meeting of the American Association of Physicists in Medicine (AAPM), July 2008.

  • C. Sagedy, N. Kandasamy, and G. Sharp, "Accelerating B-Spline Registration using Graphics Processing Units," Annual Meeting of the American Association of Physicists in Medicine (AAPM), July 2009.

  • G. Sharp, Z. Wu, and N. Kandasamy, "A Data Structure for B-Spline Registration," Annual Meeting of the American Association of Physicists in Medicine (AAPM), July 2008.

  • H. Singh, N. Kandasamy, G. Sharp, and M. Folkert, "A Streaming Architecture for Conebeam CT Image Reconstruction and Registration," The Annual Meeting of the American Association of Physicists in Medicine (AAPM), July 2007.