在应用磁共振血管造影图像进行临床诊断时，临床医生往往需要提取感兴趣区域（Region Of Interest，ROI）的部分血管．这个工作传统上需要手工进行，费时费力．该文提出一种并行的血管分割与追踪算法，利用现代图形处理器（Graphics Processing Unit，GPU）所具备的大规模并行计算能力进行快速的血管分割．首先将三维图像网格化为共面的立方体，并行处理每个立方体，确定立方体中哪些表面有血管通过，以及立方体中哪些体素包含血管．之后再将该结果用于串行的全局分割与血管追踪处理．实验结果表明，利用这种先并行后串行的方法，可以在1 s之内完成全脑血管的分割，分割的结果也更准确．
马超
,
张雪莹
,
王成龙
,
谢海滨
,
陆建平
,
杨光
,
张成秀
. 基于GPU加速的磁共振血管造影图像的并行分割与追踪算法[J]. 波谱学杂志, 2016
, 33(4)
: 570
-580
.
DOI: 10.11938/cjmr20160406
Clinical magnetic resonance angiography (MRA) often involves extraction of images, which is often done manually by radiologists. The process can be tedious and time-consuming. In this study, we propose a new parallel vessel segmentation/tracking algorithm, utilizing large-scale parallel computing provided by graphics processing unit (GPU). The whole three-dimensional image volumes are first divided into small cubes, which share surface with their neighbors. Each cube is then processed separately to determine whether there are vessels passing through its surface. These results are then used for global segmentation and vessel tracking. Application of the algorithm to real MRA data showed that segmentation of a whole-brain MRA dataset could be achieved in less than 1 s.
[1] LU Jian-ping(陆建平), LIU Qi(刘崎). Three dimensional contrast-enhanced magnetic resonance angiography(三维增强磁共振血管成像)[M]. Shanghai(上海):Shanghai Science and Technology Press(上海科学技术出版社), 2005.
[2] Yang Bao-lian(杨保联). Future of ultra high field MRI in basic research and clinical applications(超高场磁共振人体成像应用研究和医学前景)[J]. Chinese J Magn Reson(波谱学杂志), 2015, 32(4):707-714.
[3] Lesage D, Angelini E D, Bloch I, et al. A review of 3D vessel lumen segmentation techniques:Models, features and extraction schemes[J]. Med Imag Anal, 2009, 13(6):819-845.
[4] Lorensen W E, Cline H E. Marching cubes:A high resolution 3D surface construction algorithm[C]//ACM siggraph computer graphics. ACM, 1987, 21(4):163-169.
[5] Roshnara Nasrin P P, Jabbar S. Efficient 3D visual hull reconstruction based on marching cube algorithm[C]//Innovations in Information, Embedded and Communication Systems (ICⅡECS), 2015 International Conference on IEEE, 2015:1-6.
[6] Adams R, Bischof L. Seeded region growing[J]. IEEE T Pattern Anal, 1994, 16(6):641-647.
[7] Hojjatoleslami S A, Kittler J. Region growing:A New Approach[J]. IEEE T Image Process, 1998, 7(7):1079-1084.
[8] Rouhi R, Jafari M, Kasaei S, et al. Benign and malignant breast tumors classification based on region growing and CNN segmentation[J]. Expert Syst Appl, 2015, 42(3):990-1002.
[9] Wang Cheng-long(王成龙). A Study on Fast Vessel Segmentation and Tracking Algorithm(快速血管分割追踪算法的研究)[D]. Shanghai(上海):East China Normal University(华东师范大学), 2014.
[10] Farber R. CUDA Application Design and Development[M]. USA:Elsevier, 2011.
[11] Owens J D, Houston M, Luebke D, et al. GPU computing[J]. P IEEE, 2008, 96(5):879-899.
[12] Harish P, Narayanan P J. Accelerating Large Graph Algorithms on the GPU Using CUDA[M]. Berlin Heidelberg:Springer, 2007:197-208.
[13] Pan L, Gu L, Xu J. Implementation of medical image segmentation in CUDA[C]//Information Technology and Applications in Biomedicine, 2008. International Conference on IEEE, 2008:82-85.
[14] Smith S M, Brady J M. SUSAN-a new approach to low level image processing[J]. In J Comput Vis, 1997, 23(1):45-78.
[15] Qu Z, Wang P, Gao Y, et al. Fast SUSAN edge detector by adapting step-size[J]. Optik-International Journal for Light and Electron Optics, 2013, 124(8):747-750.
[16] Rezai-Rad G, Aghababaie M. Comparison of SUSAN and sobel edge detection in MRI images for feature extraction[C]//Information and Communication Technologies, 2006. ICTTA'06. 2nd. IEEE, 2006, 1:1103-1107.
[17] Otsu N. A threshold selection method from gray-level histograms[J]. Automatica, 1975, 11(285-296):23-27.
[18] Smistad E, Falch T L, Bozorgi M, et al. Medical image segmentation on GPUs-A comprehensive review[J]. Med Imag Anal, 2015, 20(1):1-18.
