Research project

Hardware Acceleration Techniques for 3D Cone-Beam Reconstruction

In the scope of a collaboration between the Pattern Recognition Lab, Siemens Healthcare (CV Division) and Intel Corp. research in the area of innovative developments of computed tomography (CT) and its implementation on different hardware architectures has been conducted since November 2004. The main interest lies on (iterative) algorithms for 3D-reconstruction in C- arm CT systems.

Reconstruction methods in CT require an immense computing power of the reconstruction system, because of the amount of projection data to be processed and also because of the number and size of the slice images to be computed. In order to meet the requirements for the computing time of productive systems, it is still necessary to use high performance hardware components. This project concentrates on parallel implementations of different reconstruction algorithms on multi processor systems, graphics accelerator boards (NVIDIA CUDA, Intel Larrabee), specialized processors (IBM Cell), as well as on reconfigurable accelerator hardware based on Field Programmable Gate Array (FPGA) technology.

So far the optimized implementation of iterative reconstruction algorithms and the Feldkamp method on GPUs using CUDA have been developed. Since a comparison of published results in this field is hardly possible a standardized data set and corresponding performance measurement software were released. The performance of developed highly optimized many-core CPU implementations was published on the accompanying website "www.rabbitct.com". Furthermore, the achieved insights were transferred to Intel's new hardware architecture, Larrabee, and extended by techniques to exploit its novel features.

Project manager:
Prof. Dr.-Ing. Joachim Hornegger, Dr.-Ing. Markus Kowarschik

Project participants:
Dr.-Ing. Benjamin Keck, Dipl.-Inf. Hannes Hofmann, Dr.-Ing. Holger Scherl

Keywords:
Computed tomography; 3D-reconstruction; Hardware programming; Parallel Algorithms; Performance Evaluation; CUDA; Larrabee; GPU

Duration: 1.11.2004 - 31.3.2011

Sponsored by:
Siemens AG Healthcare Sector (Components & Vacuum Technology)
Intel Corporation

Contact:
Keck, Benjamin
E-Mail: keck@informatik.uni-erlangen.de

Publications

Keck, Benjamin: Medical Image Reconstruction using Graphics Hardware (CUDA).Talk: Gastvortrag, Radiological Sciences Laboratory, Stanford University School of Medicine, Stanford University, CA, USA, 11.11.2009

Keck, Benjamin ; Hofmann, Hannes ; Scherl, Holger ; Kowarschik, Markus ; Hornegger, Joachim: High Resolution Iterative CT Reconstruction using Graphics Hardware. In: Yu, Bo (Ed.) : 2009 IEEE Nuclear Science Symposium Conference Record (IEEE Medical Imaging Conference 2009 Orlando, USA - 31 Oktober 2009). N/A : Omnipress, 2009, pp 4035-4040.

Hofmann, Hannes ; Keck, Benjamin ; Rohkohl, Christopher ; Hornegger, Joachim: Towards C-arm CT Reconstruction on Larrabee. In: Tsui, Benjamin M. W. (Ed.) : Proceedings of 10th Fully 3D Meeting and 2nd HPIR Workshop (10th Fully 3D Meeting and 2nd HPIR Workshop Beijing 5-10.9.2009). 2009, pp 1-4.

Rohkohl, Christopher ; Keck, Benjamin ; Hofmann, Hannes ; Hornegger, Joachim: RabbitCT - an open platform for benchmarking 3D cone-beam reconstruction algorithms. In: Medical Physics 36 (2009), No. 9, pp 3940-3944
[doi>10.1118/1.3180956]

Keck, Benjamin ; Hofmann, Hannes ; Scherl, Holger ; Kowarschik, Markus ; Hornegger, Joachim: GPU-accelerated SART reconstruction using the CUDA programming environment. In: Samei, Ehsan ; Hsieh, Jiang (Ed.) : Proceedings of SPIE (Medical Imaging 2009: Physics of Medical Imaging Lake Buena Vista 7-12.2.2009). Vol. 7258, 1. Edition Lake Buena Vista : SPIE, 2009, pp 72582B.

Scherl, Holger ; Keck, Benjamin ; Kowarschik, Markus ; Hornegger, Joachim: Fast GPU-Based CT Reconstruction using the Common Unified Device Architecture (CUDA). In: Frey, Eric C. (Ed.) : Nuclear Science Symposium, Medical Imaging Conference 2007 (2007 Nuclear Science Symposium, Medical Imaging Conference Honolulu, Hawaii (USA) 30.10. - 3.11.2007). Vol. 6. 2007, pp 4464-4466. (Nuclear Science Symposium Conference Record, 2007. NSS '07. IEEE) - ISBN 978-1-4244-0922-8

Scherl, Holger ; Hoppe, Stefan ; Dennerlein, Frank ; Lauritsch, Günter ; Eckert, Wieland ; Kowarschik, Markus ; Hornegger, Joachim: On-the-fly-Reconstruction in Exact Cone-Beam CT using the Cell Broadband Engine Architecture. In: .. (Ed.) : Proceedings Fully3D Meeting and HPIR Workshop (9th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine Lindau July 9 - 13, 2007). 2007, pp 29-32.

Scherl, Holger ; Koerner, Mario ; Hofmann, Hannes ; Eckert, Wieland ; Kowarschik, Markus ; Hornegger, Joachim: Implementation of the FDK Algorithm for Cone-Beam CT on the Cell Broadband Engine Architecture. In: Hsieh, J. ; Flynn, M. J. (Ed.) : Proceedings of SPIE (SPIE Medical Imaging - Physics of Medical Imaging San Diego 17-22.2.2007). Vol. 6510. 2007, pp 651058.

Weinlich, Andreas ; Keck, Benjamin ; Scherl, Holger ; Kowarschik, Markus ; Hornegger, Joachim: Comparison of High-Speed Ray Casting on GPU using CUDA and OpenGL. In: Buchty, Rainer ; Weiß, Jan-Philipp (Ed.) : Proceedings of the First International Workshop on New Frontiers in High-performance and Hardware-aware Computing (HipHaC'08) (First International Workshop on New Frontiers in High-performance and Hardware-aware Computing (HipHaC'08) Lake Como, Italy 8.11.2008). Vol. 1, 1. Edition Karlsruh : Universitätsverlag Karlsruhe, 2008, pp 25-30. - ISBN 978-3-86644-298-6

Hofmann, Hannes: Fast CT Reconstruction Using a Standardized Benchmark.Talk: 22. PARS - Workshop, GI/ITG-Fachgruppe 'Parallel-Algorithmen, -Rechnerstrukturen und -Systemsoftware (PARS)', Parsberg in der Oberpfalz, 5.6..2009

Bührle, Elmar ; Keck, Benjamin ; Böhm, Stefan ; Hornegger, Joachim: Mehrstufige zeit- und bewegungsabhängige Rauschreduktion in Echtzeit mittels CUDA. In: Meinzer, Hans-Peter ; Deserno, Thomas Martin ; Handels, Heinz ; Tolxdorff, Thomas (Ed.) : Bildverarbeitung für die Medizin 2009 - Algorithmen Systeme Anwendungen (Bildverarbeitung für die Medizin 2009 - Algorithmen Systeme Anwendungen Heidelberg 22-25.4.2009). Vol. 1. Heidelberg : Springer, 2009, pp 464-468. - ISBN 978-3-540-93859-0

Scherl, Holger ; Kowarschik, Markus ; Hornegger, Joachim: Bit-Accurate Simulation of Convolution-Based Filtering on Reconfigurable Hardware. In: Hülsemann, Frank ; Kowarschik, Markus ; Rüde, Ulrich (Ed.) : Frontiers in Simulation (Simulationstechnique 18th Symposium in Erlangen 2005 Erlangen 12.-15. September 2005). Erlangen : SCS Publishing House e.V., 2005, pp 662-667. - ISBN 3-936150-41-9