基于GPU硬件加速的医学图像融合研究
|
摘要
医学图像融合是目前的一个研究热点问题,是一个多学科交叉的研究领域,是计算机图形学和图象处理在生物医学工程中的重要应用。它涉及数字图象处理、计算机图形学以及医学领域的相关知识。医学图像融合及可视化是一种保存不同成像模式图像的特征,将其融合在一张图像上显示的图形图像处理技术,它能够反应病变区域空间位置对应关系,以及综合评价病人病理特性,为更有依据的医疗诊断及后续治疗起了相当大的帮助。它在诊断医学、手术规划及模拟仿真、整形及假肢外科、放射治疗规划、解剖教学等应用方向都发挥了重要作用。因此,对医学图像融合的研究,具有重要的学术意义和应用价值。然而过去的图像融合技术,在精度以及计算时间上都不够理想,从而在一些方面上限制了其的应用范围。
随着计算机硬件的高速发展,图形处理器(GPU)的计算能力也保持着几何级数的增长趋势,不仅被运用于图形渲染,还凭借着它的出色的浮点计算能力、灵活的可编程性以及并行架构,被越来越多的应于图形学之外的其他通用计算领域,从而形成了一项新的技术——通用图形硬件加速编程(GPGPU)技术。
本文研究的主要内容是提出了一种新的多尺度算法——频域非下采样轮廓波变换,从理论上证明了该变换算法,并且研究了GPU在与医学图像分割相关的图形计算和通用计算中的应用技术,包括利用GPU的图形渲染管线进行通用计算的基本思路和技巧,最后将GPU加速与多尺度变换相结合,介绍了基于GPU实现融合算法的具体步骤,并通过实验证明了GPU对于医学图像融合所具有的巨大加速作用。
Medical image fusion is a multi-disciplinary subject and an important application of digital image processing in biomedical engineering that relates to image and signal processing as well as medical imaging and clinical diagnosis. This subject combines distinguished characters of different data sources into one image, and represents their relationship of space to help doctors to diagnose patents’illnesses and arrange their post hospitalization treatments more precisely and comprehensively. Fusion and visualization of medical images plays an important role in diagnosis, surgery planning and simulating, radiotherapy planning and teaching in anatomy. Thus, study and research on the medical image fusion have significance on science and worthiness in practical application. However, since medical image fusion technical has not done well in both accuracy and speed to restrict its field of application.
With the advance in computer hardware, the computation power of graphics processing unit (GPU) is growing exponentially in the latest several years. Today its applied range is not only limited in graphics rendering, but also extends to some general purpose computing by its excellent float-point arithmetic capacity, flexible programming and parallel architecture. This is led to the so-called GPGPU technology.
The main subject of this thesis is to propose and demonstrate a new multiple dimensioned arithmetic named frequency domain Nonsubsampled Contourlet transform. And apply this arithmetic to medical image fusion and accelerate the whole process by GPU. This algorithm builds a simple and high-speed Nonsubsampled Contourlet transform in frequency domain to eliminate the ringing and Pseudo-Gibbs phenomena caused by Wavelet or Contourlet Transformation. The thesis has also done some research on the application technology of medical image visualization, fusion, graphics and general purpose computing. Finally, the thesis introduces the specific image fusion process and compare the performances between different platforms and algorithms, which indicates that this fusion method can provide great definition and acceleration for medical image fusion.
随着计算机硬件的高速发展,图形处理器(GPU)的计算能力也保持着几何级数的增长趋势,不仅被运用于图形渲染,还凭借着它的出色的浮点计算能力、灵活的可编程性以及并行架构,被越来越多的应于图形学之外的其他通用计算领域,从而形成了一项新的技术——通用图形硬件加速编程(GPGPU)技术。
本文研究的主要内容是提出了一种新的多尺度算法——频域非下采样轮廓波变换,从理论上证明了该变换算法,并且研究了GPU在与医学图像分割相关的图形计算和通用计算中的应用技术,包括利用GPU的图形渲染管线进行通用计算的基本思路和技巧,最后将GPU加速与多尺度变换相结合,介绍了基于GPU实现融合算法的具体步骤,并通过实验证明了GPU对于医学图像融合所具有的巨大加速作用。
Medical image fusion is a multi-disciplinary subject and an important application of digital image processing in biomedical engineering that relates to image and signal processing as well as medical imaging and clinical diagnosis. This subject combines distinguished characters of different data sources into one image, and represents their relationship of space to help doctors to diagnose patents’illnesses and arrange their post hospitalization treatments more precisely and comprehensively. Fusion and visualization of medical images plays an important role in diagnosis, surgery planning and simulating, radiotherapy planning and teaching in anatomy. Thus, study and research on the medical image fusion have significance on science and worthiness in practical application. However, since medical image fusion technical has not done well in both accuracy and speed to restrict its field of application.
With the advance in computer hardware, the computation power of graphics processing unit (GPU) is growing exponentially in the latest several years. Today its applied range is not only limited in graphics rendering, but also extends to some general purpose computing by its excellent float-point arithmetic capacity, flexible programming and parallel architecture. This is led to the so-called GPGPU technology.
The main subject of this thesis is to propose and demonstrate a new multiple dimensioned arithmetic named frequency domain Nonsubsampled Contourlet transform. And apply this arithmetic to medical image fusion and accelerate the whole process by GPU. This algorithm builds a simple and high-speed Nonsubsampled Contourlet transform in frequency domain to eliminate the ringing and Pseudo-Gibbs phenomena caused by Wavelet or Contourlet Transformation. The thesis has also done some research on the application technology of medical image visualization, fusion, graphics and general purpose computing. Finally, the thesis introduces the specific image fusion process and compare the performances between different platforms and algorithms, which indicates that this fusion method can provide great definition and acceleration for medical image fusion.
引文
[1]王江涛,医学图像融合的临床应用与新进展,武汉,医学影像杂,19(4), 2009:476-478.
[2] Maintz J.B.A, Viergever M.A. A survey of medical image registration. Medical Image Analysis,1998,1:1-36.
[3]王梦倩,邵波,岳建华。医学图像融合的应用与展望。医疗设备信息,2007,5:48-50.
[4] Sannazzari G.L, Ragona R, Redda M.G. CT-MRI image fusion for delineation of volumes in three-dimensional conformal radiation therapy in the treatment of localized prostate cancer. British Journal of Radiology, 2002, 75:603-607
[5] Magnani P, Carretta A, Rizzo G, et al. FDG/PET and spiral CT image fusion for medistinal lymph node assessment of non-small cell lung cancer patients. J Cardiovasc Surg(Turino),1999,40:741-748
[6]汪家旺,罗立民,舒华忠。CT、MR图像融合技术临床应用研究。中华放射学杂志,2001,8:604-608.
[7]李坤成,李永忠,赵富强。CT与MRI图像配准与融合技术在颅脑肿瘤中的初步应用。中国医学影像技术,2001,17:823-824
[8] Von Schmethess G.K, Steinert H.C, Hang T.F, et al. Integrated PET/CT: current applications and future directions. Radiology, 2006, 238(2):405-422.
[9] Goerres G.W, Von Schulthess G.K, Steinert H.C, et al. Why most PET of lung and head-and-neck cancer will be PET/CT. Journal of Nuclear Medicine, 2004, 45(1):55-71.
[10]姚原,吴国华,夏士安等。CT/MRI影像融合对鼻咽癌肿瘤靶区以及三维适形治疗的影响。中国癌症杂志,2—6,16(6):498-502
[11] Petra A, Elsen V. Medical image matching– A review with classification. IEEE Transactions on Medical Image,1993,12(3):26-39.
[12]李晓光,谭建豪。基于互信息的磁共振颅脑图像配准,2008,25(12):216-219
[13] T.Pu and G.Ni. Contrast-based image fusion using the discrete wavelet transform. Opt. Eng., 2000, 39(8): 2075-2082.
[14] Z.Zhang, R. Blum. A categorization of multiscale-decomposition-based image fusion schemes with a performance study for a digital camera application. Proceedings of IEEE, USA, Aug.1999, 87(8): 1315-1326.
[15] Mrazek P, Weickert J, Steidl G. Diffusion-Inspired Shrinkage Functions and Stability Results for Wavelet Denoising. 2005,64(3):171-186.
[16] Do M.N, Vetterli M. The contourlet transform: an efficient directional multiresolution image representation. IEEE Transaction on Image Processing, 2005, 14(12): 2091-2106.
[17] Cunha A.L, Zhou J. The Nonsubsampled Contourlet Transform: Theory, Design, and Applications. IEEE Transaction on Image Processing, Oct 2006, 15(10): 3089-3101.
[18] Owens J.D, Luebke, D, Govindaraju, N., etc. A survey of general-purpose computation on graphics hardware. Computer Graphics Forum. 2007, 26(1): 80-113.
[19] From Wikipedia. Graphics processing unit. http://en.wikipedia.org/wiki/Graphics_processing _unit. 2009.
[20] Kruger J, Westermann R. Linear algebra operators for GPU implementation of numerical algorithms. ACM Transactions on Graphics. 2003, 22(3): 908–916.
[21] Bolz J, Farmer I, Grinspun E, Schr?der P. Sparse matrix solvers on the GPU: Conjugate gradients and multi-grid. ACM Transactions on Graphics. 2003, 22(3): 917–924.
[22] Kold A, Cuntz N. Dynamic particle coupling for GPU-based fluid simulation. Proceedings of the 18th Symposium on Simulation Technique. Germany. 2005: 722–727.
[23] Bond A. Havok FX: GPU-accelerated physics for PC games. Proceedings of Game Developers Conference 2006. USA. 2006
[24] Fung J., Mann S. Computer vision signal processing on graphics processing units. Proceedings of the IEEE International Conference on Acoustics. Speech, and Signal Processing. Canada. 2004: 93–96.
[25] Farrugia J.P, Horain P, Guehenneux E, et al. GPUCV: A framework for image processing acceleration with graphics processors. Proceedings of 2006 IEEE International Conference on Multimedia and Expo. Canada. 2006: 585-588.
[26] Jodoin P.M, Mignotte M. Markovian segmentation and parameter estimation on graphics hardware. Journal of Electrical Imaging. 2006, 15(3): 033005-1– 033005-15.
[27] Lefohn A.E, Kniss J.M, Hansen C.D, etc. A streaming narrow-band algorithm: interactive computation and visualization of level sets. IEEE Transactions on Visualization and Computer Graphics. 2004, 10(4): 422-433.
[28] M.N.Do, M.Vetterli. The contourlet transform: an efficient directional multiresolution image representation. IEEE Transaction on Image Processing, 2005, 14(12): 2091-2106.
[29] Shensa M.J. The discrete wavelet transform: Wedding the a trous and Mallat algorithms. IEEE Trans. Signal Process, 1992, 40(10): 2464-2482.
[30] Burt P.J, Adelson E.H. The Laplacian pyramid as a compact image code. IEEE Transaction on Communications, 1983,31(4):533-542.
[31] Bamberger R.H. and Smith M.J. A filter bank for the directional decomposition of images: Theory and design. IEEE Trans.Signal Process, 1992, 40(4): 882-893.
[32] Cunha A. L, Zhou J. The Nonsubsampled Contourlet Transform: Theory, Design, and Applications. IEEE Transaction on Image Processing, Oct 2006, 15(10): 3089-3101.
[33] J.H.McClellan. The design of two-dimensional digital filters by transformation. Proc. 7th Annu. Princeton Conf on Information Sciences and Systems, 1973, 247-251.
[34] Randi J.,天宏工作室译. OpenGL着色语言.北京.人民邮电出版社.2006:1-23
[35]韦轶群.基于GPU硬件加速的医学图像分割研究.[学位论文].上海.上海交通大学.2009:34-45
[36] D. G?ddeke. GPGPU Tutorials[EB]. http://www.mathematik.uni-dortmund.de/. 2003.
[37] Pharr, M.,龚敏敏译. GPU精粹2——高性能图形芯片和通用计算编程技巧.北京:清华大学出版社. 2007:331-422.
[38] OpenGL Extension Registry. http://www.opengl.org/registry/, 2008.
[39]冯煌. GPU图像处理的FFT和卷积算法性能分析.计算机工程与运用. 2008,44(2):120-122
[40]Micheli G., Gipta R. K.. Hardware/software co-design. Proceedings of the IEEE, 1997(3):349-365.
[41]秦红星.基于点模型的体绘制和3D几何滤波研究. [学位论文].上海.上海交通大学. 2008: 9-10.
[42] Lorensen, E.W., Cline, H.E. Marching Cubes: A high resolution 3D surface construction algorithm. Computer Graphics. 1987, 21(4): 163-169.
[43]唐炜.基于GPU的医学图像高质量体绘制技术.[学位论文].上海.上海交通大学.2008: 13-60, 70-77.
[44] Akimoto, T., Mase, K., Suenaga,Y. Pixel-selected ray tracing. IEEE Computer Graphics andApplication, 11(4), 1991: 14-22.
[45]孙志刚,张加万.一种改进的Splatting体绘制方法.天津大学学报. 2003, 36(5): 626-630.
[46] Lacroute, P., Levoy M. Fast volume rendering using a shear-warp factorization of the viewing transformation. In Proceedings of the 21st annual conference on Computer graphics and interactive techniques. USA. 1994: 451-458.
[47] Deng, J., Tang, Z. Color assignment and boundary sharpening in frequency domain volume rendering. In Proceeding of Pacific Graphics’95. Korea. 1995: 240-252.
[48] Michel, A.W., Roerdink, Jos B.T.M. X-ray volume rendering by hierarchical wavelet splatting. In Proceedings of the International Conference on Pattern Recognition. USA. 2000: 3163.
[49] Tatarchuk, N., Shopf, J. Real-time medical visualization with FireGL. SIGGRAPH 2007. USA. 2007.
[50]Matsui M.,Ino F.,Hagihara K., Parallel volume rendering with early ray termination for visualizing large-scale datasets, ISPA 2004, 2004:245-256.
[51]Li W.,Mueller K.,Kaufman A., Empty space skipping and occlusion clipping for texture-based volume rendering. IEEE Visualization(2003):317-324.
[52] Levoy M., Efficient raytracing of volume data, ACM Transactions on Graphics, vol.9(3), 1990:245-261.
[53]童欣等,基于空间跳跃的三维纹理硬件体绘制算法,计算机学报, Vol.21(9), 1999:807-812.
[54] Shreiner, D., Woo, M., Neider, J.,徐波等译. OpenGL编程指南(第5版).北京机械工业出版社. 2006: 140-145.
[55]吴佳鹏,杨兆选,苏育挺,等.基于模糊推理的小波域图像融合规则.天津大学学报,2007,40(12).
[56] Burt P.J. and Koiczynski R.J. Enhancement with application to image fusion[C]. Proc of the 4th Int. Conf. on Computer Vision., Los. Alamitos: IEEE Computer Society, 1993. 173-182.
[57]郭雷,等.图像融合.北京:电子工业出版社, 2008.
[2] Maintz J.B.A, Viergever M.A. A survey of medical image registration. Medical Image Analysis,1998,1:1-36.
[3]王梦倩,邵波,岳建华。医学图像融合的应用与展望。医疗设备信息,2007,5:48-50.
[4] Sannazzari G.L, Ragona R, Redda M.G. CT-MRI image fusion for delineation of volumes in three-dimensional conformal radiation therapy in the treatment of localized prostate cancer. British Journal of Radiology, 2002, 75:603-607
[5] Magnani P, Carretta A, Rizzo G, et al. FDG/PET and spiral CT image fusion for medistinal lymph node assessment of non-small cell lung cancer patients. J Cardiovasc Surg(Turino),1999,40:741-748
[6]汪家旺,罗立民,舒华忠。CT、MR图像融合技术临床应用研究。中华放射学杂志,2001,8:604-608.
[7]李坤成,李永忠,赵富强。CT与MRI图像配准与融合技术在颅脑肿瘤中的初步应用。中国医学影像技术,2001,17:823-824
[8] Von Schmethess G.K, Steinert H.C, Hang T.F, et al. Integrated PET/CT: current applications and future directions. Radiology, 2006, 238(2):405-422.
[9] Goerres G.W, Von Schulthess G.K, Steinert H.C, et al. Why most PET of lung and head-and-neck cancer will be PET/CT. Journal of Nuclear Medicine, 2004, 45(1):55-71.
[10]姚原,吴国华,夏士安等。CT/MRI影像融合对鼻咽癌肿瘤靶区以及三维适形治疗的影响。中国癌症杂志,2—6,16(6):498-502
[11] Petra A, Elsen V. Medical image matching– A review with classification. IEEE Transactions on Medical Image,1993,12(3):26-39.
[12]李晓光,谭建豪。基于互信息的磁共振颅脑图像配准,2008,25(12):216-219
[13] T.Pu and G.Ni. Contrast-based image fusion using the discrete wavelet transform. Opt. Eng., 2000, 39(8): 2075-2082.
[14] Z.Zhang, R. Blum. A categorization of multiscale-decomposition-based image fusion schemes with a performance study for a digital camera application. Proceedings of IEEE, USA, Aug.1999, 87(8): 1315-1326.
[15] Mrazek P, Weickert J, Steidl G. Diffusion-Inspired Shrinkage Functions and Stability Results for Wavelet Denoising. 2005,64(3):171-186.
[16] Do M.N, Vetterli M. The contourlet transform: an efficient directional multiresolution image representation. IEEE Transaction on Image Processing, 2005, 14(12): 2091-2106.
[17] Cunha A.L, Zhou J. The Nonsubsampled Contourlet Transform: Theory, Design, and Applications. IEEE Transaction on Image Processing, Oct 2006, 15(10): 3089-3101.
[18] Owens J.D, Luebke, D, Govindaraju, N., etc. A survey of general-purpose computation on graphics hardware. Computer Graphics Forum. 2007, 26(1): 80-113.
[19] From Wikipedia. Graphics processing unit. http://en.wikipedia.org/wiki/Graphics_processing _unit. 2009.
[20] Kruger J, Westermann R. Linear algebra operators for GPU implementation of numerical algorithms. ACM Transactions on Graphics. 2003, 22(3): 908–916.
[21] Bolz J, Farmer I, Grinspun E, Schr?der P. Sparse matrix solvers on the GPU: Conjugate gradients and multi-grid. ACM Transactions on Graphics. 2003, 22(3): 917–924.
[22] Kold A, Cuntz N. Dynamic particle coupling for GPU-based fluid simulation. Proceedings of the 18th Symposium on Simulation Technique. Germany. 2005: 722–727.
[23] Bond A. Havok FX: GPU-accelerated physics for PC games. Proceedings of Game Developers Conference 2006. USA. 2006
[24] Fung J., Mann S. Computer vision signal processing on graphics processing units. Proceedings of the IEEE International Conference on Acoustics. Speech, and Signal Processing. Canada. 2004: 93–96.
[25] Farrugia J.P, Horain P, Guehenneux E, et al. GPUCV: A framework for image processing acceleration with graphics processors. Proceedings of 2006 IEEE International Conference on Multimedia and Expo. Canada. 2006: 585-588.
[26] Jodoin P.M, Mignotte M. Markovian segmentation and parameter estimation on graphics hardware. Journal of Electrical Imaging. 2006, 15(3): 033005-1– 033005-15.
[27] Lefohn A.E, Kniss J.M, Hansen C.D, etc. A streaming narrow-band algorithm: interactive computation and visualization of level sets. IEEE Transactions on Visualization and Computer Graphics. 2004, 10(4): 422-433.
[28] M.N.Do, M.Vetterli. The contourlet transform: an efficient directional multiresolution image representation. IEEE Transaction on Image Processing, 2005, 14(12): 2091-2106.
[29] Shensa M.J. The discrete wavelet transform: Wedding the a trous and Mallat algorithms. IEEE Trans. Signal Process, 1992, 40(10): 2464-2482.
[30] Burt P.J, Adelson E.H. The Laplacian pyramid as a compact image code. IEEE Transaction on Communications, 1983,31(4):533-542.
[31] Bamberger R.H. and Smith M.J. A filter bank for the directional decomposition of images: Theory and design. IEEE Trans.Signal Process, 1992, 40(4): 882-893.
[32] Cunha A. L, Zhou J. The Nonsubsampled Contourlet Transform: Theory, Design, and Applications. IEEE Transaction on Image Processing, Oct 2006, 15(10): 3089-3101.
[33] J.H.McClellan. The design of two-dimensional digital filters by transformation. Proc. 7th Annu. Princeton Conf on Information Sciences and Systems, 1973, 247-251.
[34] Randi J.,天宏工作室译. OpenGL着色语言.北京.人民邮电出版社.2006:1-23
[35]韦轶群.基于GPU硬件加速的医学图像分割研究.[学位论文].上海.上海交通大学.2009:34-45
[36] D. G?ddeke. GPGPU Tutorials[EB]. http://www.mathematik.uni-dortmund.de/. 2003.
[37] Pharr, M.,龚敏敏译. GPU精粹2——高性能图形芯片和通用计算编程技巧.北京:清华大学出版社. 2007:331-422.
[38] OpenGL Extension Registry. http://www.opengl.org/registry/, 2008.
[39]冯煌. GPU图像处理的FFT和卷积算法性能分析.计算机工程与运用. 2008,44(2):120-122
[40]Micheli G., Gipta R. K.. Hardware/software co-design. Proceedings of the IEEE, 1997(3):349-365.
[41]秦红星.基于点模型的体绘制和3D几何滤波研究. [学位论文].上海.上海交通大学. 2008: 9-10.
[42] Lorensen, E.W., Cline, H.E. Marching Cubes: A high resolution 3D surface construction algorithm. Computer Graphics. 1987, 21(4): 163-169.
[43]唐炜.基于GPU的医学图像高质量体绘制技术.[学位论文].上海.上海交通大学.2008: 13-60, 70-77.
[44] Akimoto, T., Mase, K., Suenaga,Y. Pixel-selected ray tracing. IEEE Computer Graphics andApplication, 11(4), 1991: 14-22.
[45]孙志刚,张加万.一种改进的Splatting体绘制方法.天津大学学报. 2003, 36(5): 626-630.
[46] Lacroute, P., Levoy M. Fast volume rendering using a shear-warp factorization of the viewing transformation. In Proceedings of the 21st annual conference on Computer graphics and interactive techniques. USA. 1994: 451-458.
[47] Deng, J., Tang, Z. Color assignment and boundary sharpening in frequency domain volume rendering. In Proceeding of Pacific Graphics’95. Korea. 1995: 240-252.
[48] Michel, A.W., Roerdink, Jos B.T.M. X-ray volume rendering by hierarchical wavelet splatting. In Proceedings of the International Conference on Pattern Recognition. USA. 2000: 3163.
[49] Tatarchuk, N., Shopf, J. Real-time medical visualization with FireGL. SIGGRAPH 2007. USA. 2007.
[50]Matsui M.,Ino F.,Hagihara K., Parallel volume rendering with early ray termination for visualizing large-scale datasets, ISPA 2004, 2004:245-256.
[51]Li W.,Mueller K.,Kaufman A., Empty space skipping and occlusion clipping for texture-based volume rendering. IEEE Visualization(2003):317-324.
[52] Levoy M., Efficient raytracing of volume data, ACM Transactions on Graphics, vol.9(3), 1990:245-261.
[53]童欣等,基于空间跳跃的三维纹理硬件体绘制算法,计算机学报, Vol.21(9), 1999:807-812.
[54] Shreiner, D., Woo, M., Neider, J.,徐波等译. OpenGL编程指南(第5版).北京机械工业出版社. 2006: 140-145.
[55]吴佳鹏,杨兆选,苏育挺,等.基于模糊推理的小波域图像融合规则.天津大学学报,2007,40(12).
[56] Burt P.J. and Koiczynski R.J. Enhancement with application to image fusion[C]. Proc of the 4th Int. Conf. on Computer Vision., Los. Alamitos: IEEE Computer Society, 1993. 173-182.
[57]郭雷,等.图像融合.北京:电子工业出版社, 2008.