摘要
心脏核磁共振成像(magnetic resonance imaging, MRI)能够提供高分辨率、高品质的医学图像,是当前医学图像分析领域的研究热点之一。借助于心脏核磁共振图像,评定左心室心肌功能,在临床诊断中具有重要意义。对心脏解剖结构的准确描述和功能的定量分析是以准确分割心肌的内外边缘为基础的,尤其是心脏左心室内、外膜的分割。主动轮廓模型对心脏核磁共振图像的分割有良好的效果从而在医学图像处理中得到广泛应用,本文研究重点正是基于Snake模型的心脏核磁共振图像分割,对外力模型的改善、内外膜分割等关键问题进行了分析和研究。
受图像质量的影响,对心脏核磁共振图像左心室内、外膜分割在有些情况下得不到很好的分割结果。本文根据目标的先验知识,引入形状约束,从而有效克服了由于图像灰度不均、乳突肌等引起的局部极小对心脏MRI的影响,使得Snake轮廓可以准确地分割心脏左心室。
本文提出了一种基于广义法向有偏梯度矢量流(GNBGVF)模型的心脏核磁共振图像左心室内、外膜分割方法。首先提出了主动轮廓模型的广义法向有偏梯度矢量流外力模型,作为对梯度矢量流的改进,该外力模型同时保持了切线方向和法线方向有偏的扩散。其次,根据左心室近似为圆形的形状特点,引入了圆形能量约束,有利于克服心脏核磁共振图像分割难点。
为了更加有效的消除噪声对图像分割的影响,本文提出了一种新的外力模型:基于扩展邻域和噪声平滑的广义梯度矢量流(ENGGVF)模型,并采用这一外力模型来分割左心室内、外膜。该模型在梯度矢量流的基础上,引入了权重因子和扩展邻域卷积运算,还在拉普拉斯算子模板中加入噪声平滑模板。同时考虑到心脏在舒张期内外膜更接近于椭圆的特点,在Snake模型中添增椭圆形状能量约束。实验表明,该方法能够快速高效准确地分割左心室内、外膜。
Cardiac magnetic resonance imaging (MRI) can provide high- resolution and high-quality medical images, which has been a hot topic in the community of medical images analysis. Cardiac magnetic resonance images can help to estimate the myocardium function of ventricles for clinical diagnosis. In order to make thorough use of the anatomical and functional information derived from images, the epicardium and endocardium of the left ventricle should be extracted. Since active contour (Snake) model has a great effect on segmentation of cardiac MR images, it has been widely used in medical image processing. This dissertation focuses on the segmentation of cardiac MR images based on snake model, which contains improved external models,segmentation of the left ventricle, etc.
To acquire satisfactory segmentation of the epicardium and endocardium of the left ventricle cardiac MR images, an energy constraint about shape based on a priori knowledge of the target is adopted. The shape constraint can conquer the unexpected local minimum stemming form image inhomogeneity and papillary muscle. With this constraint, the Snake contour is reactivated to locate the left ventricle accurately.
A novel method based on generalized normally biased gradient vector flow (GNBGVF) snake model is proposed to segment the left ventricle cardiac MR images. It first proposes an external force for active contours, which is called as generalized normally biased GVF (GNBGVF). As an improvement on gradient vector flow, the GNBGVF external force keeps the diffusion along the tangential direction of the isophotes and biases that along the normal direction simultaneously. Considering that the left ventricle is roughly a circle, a shape constraint based on circle is adopted, which conquers the difficulties in the process of segmenting cardiac MR images.
To smooth noise effectively, a novel external force called extended-neighborhood and noise-smoothing generalized gradient vector flow (ENGGVF) is proposed for segmentation of the left ventricle from cardiac MR images. The external force incorporates weighting functions and convolution operation with extended neighborhood and modifies the Laplacian operator mask by adding the noise-smoothing mask. Based on ENGGVF, the shape of the left ventricle is taken into account and an elliptic shape-based energy constraint for snake model is adopted. The proposed strategy is validated on a large amount of cardiac magnetic resonance images.
引文
[1] Ameriena Heart Association. Heart disease and stroke statistics [EB/OL]. 2008, website: http://www.americanheart.org/presenter.jhtml?identifier=1200000.
[2] Frangi A F, Niessen W J, et al. Three dimensional modeling for functional analysis of cardiac Images: A review [J]. IEEE trans Medical Imaging, 2011, 20(1):2-25.
[3] Alertras A H, et al. Dense displacement encording with simulated echoes in cardiac functional MRI [J]. Journal of Magnetic Resonance. 1999, 137(1): 247-252.
[4] Pipe J G. Method for mesduring three-dimensional motion with tagged MR Imaging [J]. Radiology, 1991, 181: 591-595.
[5] Zerhouni E A, Parish D, Rogers W, et al. Human heart: Tagging with MR imaging– a method for noninvasive assessment of myocardial motion [J]. Radiology, 1988, 169:59-63.
[6] Axel L, Dougherty L. MR imaging of motion with spatial modulation of magnetization [J]. Radiology, 1989, 171:841-845.
[7] Axel L. Heart wall motion: Improved method of spatial modulation of magnetization [J]. Radiology, 1989, 172:349—350.
[8]林瑶,田捷.医学图像分割方法综述[J].模式识别与人工智能, 2002,15(2):193-204
[9] Pham DL, Xu CY, Prince J. A survey of current methods in medical image segmentation. Technical report JHU/ECE-99-01, Johns Hopkins University, 1998.
[10] Suri JS. Computer Vision, pattern recognition and image processing in left ventricle segmentation: the last 50 years. Pattern Analysis & Application, 2000, 3:209-242.
[11] Ayache N. Medical computer vision, virtual reality and robotics. Image and Vision Computing, 1995, 13(4):295-313.
[12] Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain 3-Dimensional proportional system: an approach to cerebral imaging. Thieme medical pulisher, Inc. Stuttgart, NY, 1988.
[13] Chen SY, Lin WC, Chen CT. Split-and-merge image segmentation based on localized feature analysis and statistical tests. Graphical models and image processing, 1991, 53(5):457-475.
[14] Wu X. Adaptive split-and-merge segmentation based on piecewise least-square approximation. IEEE trans PAMI, 1993, 15(8):808-815.
[15] Haris K, Efstratiadis SN, Maglaveras N, et al. Hybrid image segmentation using watershed and fast region merging. IEEE trans Image Proc, 1998, 7(12):1684-1699.
[16] Hernandez SE, Barner KE. Joint region merging criteria for watershed-based image segmentation[C]. Proceedings of IEEE, 2000, 2:108-111.
[17] Pham DL, Xu CY, Prince J. A survey of current methods in medical image segmentation. Technical report JHU/ECE-99-01, Johns Hopkins University, 1998.
[18] Ray N, Acton ST. Image segmentation by curve evolution with clustering[C]. the 34th Asilomar Conference on Signals, Systems and Computers, 2000,1:495– 498.
[19] Leung SH, Lau WH. Lip image segmentation using fuzzy clustering incorporating an elliptic shape function. IEEE trans Image Processing, 2004,13(1):51-62.
[20] Liew AW, Leung SH, Lau WH. Segmentation of color lip images by spatial fuzzy clustering. IEEE trans Fuzzy system, 2003, 11(4):542 -549.
[21] Man Y, Gath I. Detection and separation of ring-shaped clusters using fuzzy clustering. IEEE PAMI, 1995, 16(8):855-861.
[22] Gath I, Hoory D. Fuzzy clustering of elliptic ring-shaped clusters [J]. Pattern Recognition letters, 1995, 16(7):727-741.
[23] Liew A, Yan H. An adaptive spatial fuzzy clustering algorithm for 3D MR image segmentation [J]. IEEE trans Medical Imaging, 2003, 22(9):1063 -1074.
[24] Liang Z, Macfall JR, Harrington DP. Parameter estimation and tissue segmentation from multispectral images [J]. IEEE trans Medical Imaging, 1994, 13:441-449.
[25] Geman S, Geman D. Stochastic relaxation, Gibbs distributions, and the Bayesian Restoration of images. IEEE trans PAMI, 1984, 6(6):721 -741.
[26] Pappas TN. An adaptive clustering algorithm for image segmentation [J]. IEEE trans Signal processing, 1992, 40(4):901-914.
[27] Blanz WE, Gish SL. Connectionist classifier architecture applied to image segmentation [C]. In: proceedings of the 10th int’l conf. on PR. New Jersey, USA, 1990,272-277.
[28] Babaguchi N, Yamada K, Kise K, et al. Connectionist model binarization. In: proceedings of the 10th int’l conf. on PR. New Jersey, USA, 1990, 51-56.
[29] Koenderinck J. The structure of images. Biological Cybernetics, 1984, 50:363-370.
[30] Gauch J, Oliver W, Pizer SM. Multiresolution shape descriptions and their application in medicalimaging. Proc. information processing in medical imaging, 1987, 131-149.
[31] Pizer SM, Fritsch DS. Segmentation, registration and measurement of shape variation via image object shape. IEEE trans Medical Imaging, 1999, 18(10):851-865.
[32] Tang H, Wu EX, Ma QY, et al. MARI brain image segmentation by multi-resolution edge detection and region selection. Computerized medical imaging and graphics, 2000, 24(4):349-357.
[33] Wang Y, Staib L H. Boundary finding with correspondence using statistical shape models [C]. Computer Vision and Pattern Recognition, 1998: 338-345.
[34] Hamameh G, Gustavsson T. Combining Snakes and Active Shape Models for Segmenting the Human Left Ventricle in Echocardiographic Images [C]. IEEE, Computers in Cardiology, 2000, 27: 115-118.
[35] Cootes T F, Gareth G J, Taylor C J. Active appearance models [J]. IEEE Transactions on Pattern Analysis Machine Intelligence, 2001, 23(6): 681-685.
[36] Richard W, Keen C. Automated textured based segmentation of ultrasound images of the prostate. Computerized medical imaging and graphics, 1996,20(3):131-140.
[37] Alain P, Toga A, Ayache N. Textured based MRI segmentation with a two-stage hybrid neural classifier. in Proceedings of IEEE, 2002, 2053-2057.
[38] G A Wright. Magnetic resonance imaging. IEEE Signal Processing Magazine.1997, 14(l):56-66.
[39]俎栋林.核磁共振成像学[M].北京:高等教育出版社. 2004.
[40]赵喜平.磁共振成像[M].北京:科学出版社. 2004.
[41]王元全.梯度矢量流主动轮廓模型的若干理论问题及其在医学图像分割中的应用研究[D].北京:北京理工大学,博士后研究报告, 2006年10月.
[42] Kass M, Witkin A, Terzopoulos D. Snakes: Active contour models [C]. Int'l J. computer vision, 1987, 1, 321-331.
[43] Caselles V, Catte F, Coll T, et al. A geometric model for active contours in image processing [J]. Numerische Mathematik, 1993, 66(1): 1-31.
[44] Cohen L D. On active contour models and balloons [J]. CVGIP: Image Understanding, 1991, 53(2): 211–218.
[45] Cohen L D, Cohen I, Finite-element methods for active contour models and balloons for 2-D and 3-D images [J]. IEEE Transactions on Pattern Analysis and Machine Intelligences, 1993, 15(11): 1131–1147.
[46] Xu C, Prince J. Snakes, shapes, and gradient vector flow [J]. IEEE Trans. Image Process, 1998, 7(3): 359–369.
[47] Xu C, Prince J. Generalized gradient vector flow external force for active contours [J]. Signal Processing. 1998, 71(2): 131–139.
[48] Park H K, Chung M J. External force of snake: virtual electric field [J]. Electronics Letters, 2002, 38(24), 1500-1502.
[49] Ning J, Wu C, Liu S, et al. NGVF: An improved external force field for active contour model [J]. Pattern Recognition Letters, 2007, 28:58-63.
[50] Caselles V, Morel J, Sbert C. An axiomatic approach to image interpolation [J]. IEEE Trans. on Image Processing, 1998, 7(3):376-386.
[51] Emilsson K, Kahari A, Bodin L, et al. Outer contour and radial changes of the cardiac left ventricle—A magnetic resonance imaging study. Clinical Research in Cardiology, 2007, 96(5):272-278.
[52] Caiani E G, Toledo E, MacEneaney P, et al. Automated interpretation of regional left ventricular wall motion from cardiac magnetic resonance images [J]. Journal of Cardiovascular Magnetic Resonance, 2006,8(3): 427-433
[53] Fujino T. New method of on-line quantification of regional wall motion with automated segmental motion analysis. Am. Soc. Echocardiogr. Elsevier, 2001, 14(9): 892-901.
[54] Lalande A, Jaulent M. A fuzzy automaton to detect and quantify artery lesions from arteriograms. In Proc. Information Processing and Management of Uncertainty in Knowledge Systems.Granada, Spain, 1996, 1481-1487.
[55] Cootes T F, Gareth G J, Taylor C J. Active appearance model [J]. IEEE Transactions on Pattern Analysis Machine Intelligence, 2001, 23(6): 681-685.
[56] Osher S, Sethiam J. Fronts propagating with curvature–dependent speed: Algorithms based on Hamilton-Jacobi formulation. J. Computational Phys. Elsevier, 1988, 12-49.
[57] Siddiqui K, Lauriere YB, Tannenbaum A. Area and length minimizing flows for shape segmentation.IEEE Transactions on Image Processing, 1998, 7(3):433-443.
[58] Hong H, Grosskopf S, Kim MH. Ventricular shape visualization using selective volume rendering of cardiac datasets. Computers in Biology and Medicine, 2001, -498.
[59] Makowski P, S?rensen TS, Therkildsen SV, et al. Two-Phase active contour method forsemiautomatic segmentation of the heart and blood vessels from MRI images for 3D visualization. Computerized Medical Imaging and Graphics, 2002, 26(1):9-17.
[60] Jolly MP. Automatic segmentation of the left ventricle in cardiac MR and CT images [J]. Int’l Journal of Computer Vision, 2006, 70(2): 151-163.
[61] Nguyen D, Masterson K, Vallée JP. Comparative evaluation of active contour model extensions for automated cardiac MR image segmentation by regional error assessment. Magnetic Resonance Materials in Physics, Biology and Medicine, 2007, 20(2):69-82.
[62] Nachtomy E, Coopperstein R, Vaturi M. et al. Automatic assessment of cardiac function from short-axis MRI: Procedure and clinical evaluation.MRI, 1998, 16(4): 365-376.
[63] Amol P, Uday K, Raja M. Automated Left Ventricular Segmentation in Cardiac MRI [J]. IEEE Trans BME, 2006, 53(7):1425-1428.
[64]周寿军,梁斌,陈武凡.心脏序列图像运动估计新方法:基于广义模糊梯度矢量流的形变曲线运动估计与跟踪[J].计算机学报, 2003, 26(11):1470-1478.
[65]秦安,冯前进,陈武凡. MR心脏序列图像左心室内外壁联合分割和时序追踪新方法[J].中国图象图形学报, 2008,13(1):80-88.
[66]周则明,王洪元,尤建洁, et al.基于改进快速活动轮廓模型的左心室核磁共振图像分割[J].计算机研究与发展, 2004,41(1): 136-141.
[67]王元全,贾云得.一种新的心脏核磁共振图像分割方法.计算机学报, 2007,30(1):129-136.
[68] Bertero M, Poggio T A, Torre V. Ill-posed problems in early vision. Proceedings of the IEEE, 1988, 76(8):869-889.
[69] You Y, Xu W, Tannenbaum A, et al. Behavioral analysis of anisotropic diffusion in image processing [J]. IEEE Trans. on Image Processing, 1996, 5(11):1539-1552.
[70] Wang Y, Liu L. Image segmentation using active contours with normally biased GVF external force [J]. IEEE Signal Processing Letters, 2010, 17(10):875-878.
[71] Axel L, Doughterty L. Heart wall motion: improved method of spatial modulation of magnetization for MR imaging [J]. Radiology, 1989(172):349-360.
[72]孙家广.计算机图像学[M].北京:清华大学出版社. 1998.
[73]章毓晋.图像处理[M].北京:清华大学出版社. 2005:145-149.
[74] Wang X. Laplacian Operator-Based Edge Detectors [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007, 29(5):886-890.
[75] Fitzgibbon A, Pilu M, Fisher RB. Direct least square fitting of ellipse [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1999, 21(5):476-480.