1College of Life Science and Bioengineering, Beijing University of Technology, Beijing 100124, China; 2College of Biomedical Engineering, Capital Medical University, Beijing 100069, China
Abstract：Traditional fuzzy connectedness methods exist some drawbacks in segmentation of liver vessels from computed tomography (CT) images, including unsatisfactory segmentation performance, requirement on multiple seeds, and low time efficiency. In this paper, the traditional fuzzy connectedness method was improved from following three steps: 1) The Jerman′s vesselness filter was improved; 2) The improved vesselness was incorporated into the fuzzy affinity function; 3) The fuzzy connectedness was initialized by the Otsu multi-thresholding algorithm instead of the confidence connectedness. The preprocessing comprised adaptive sigmoid filtering and isotropic resample filtering. Next, the improved Jerman′s vesselness filter was performed. Then, the improved Jerman′s vesselness was integrated into the fuzzy affinity function. The foreground information was analyzed to initialize the fuzzy connectedness by using the Otsu multi-thresholding algorithm. Finally, three-dimensional (3D) liver vessels were segmented with one single seed. The improved vesselness filter and the improved fuzzy connectedness method were quantitatively evaluated by using 20 cases of public CT data sets. The evaluation metrics included contrast to noise ratio (CNR), accuracy, sensitivity and specificity. The average CNR of the improved vesselness filter was 8.43 dB,which was superior to the traditional vesselness filters. The accuracy of the proposed vessel segmentation method was 98.11%, which was better than the traditional fuzzy connectedness method based on confidence connectedness and the regional growing and level set methods. In addition, the proposed method also had advantages in terms of time efficiency. The 3D segmentation method proposed in this paper could effectively address the issues associated with the traditional fuzzy connectedness method and improve the accuracy and efficiency of 3D liver vessel segmentation in CT images.
[1] Chen W, Zheng R, Baade PD, et al. Cancer statistics in China, 2015[J]. CA: A Cancer Journal for Clinicians, 2016, 6(2): 115-132. [2] 周莉, 陆伟. 肝癌消融治疗现状与思考[J]. 现代医药卫生, 2017, 33(3): 326-328. [3] Sonka M, Hlavac V, Boyle R. Image Processing, Analysis, and Machine vision[M] 4th edition. Stamford: Cengage Learning, 2015. [4] Sato Y, Nakajima S, Shiraga N, et al. Three-dimensional multi-scale line filter for segmentation and visualization of curvilinear structures in medical images[J]. Medical Image Analysis, 1998, 2(2): 143-168. [5] Frangi AF, Niessen WJ, Vincken KL, et al. Multiscale vessel enhancement filtering[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention.Berlin: Springer Berlin Heidelberg, 1998: 130-137. [6] Li Q, Sone S. Selective enhancement filters for nodules, vessels, and airway walls in two‐and three-dimensional CT scans[J]. Medical Physics, 2003, 30(8): 2040-2051. [7] Erdt M, Raspe M, Suehling M. Automatic hepatic vessel segmentation using graphics hardware[J]. Medical Imaging and Augmented Reality, 2008: 403-412. [8] Zhou C, Chan HP, Sahiner B, et al. Automatic multiscale enhancement and segmentation of pulmonary vessels in CT pulmonary angiography images for CAD applications[J]. Medical Physics, 2007, 34(12): 4567-4577. [9] Xiao C, Staring M, Shamonin D, et al. A strain energy filter for 3D vessel enhancement[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Berlin: Springer Berlin Heidelberg, 2010: 367-374. [10] Yang J, Ma S, Sun Q, et al. Improved Hessian multiscale enhancement filter[J]. Bio-Medical Materials and Engineering, 2014, 24(6): 3267-3275. [11] Zeng Y, Zhao Y, Tang P, et al. Liver vessel segmentation and identification based on oriented flux symmetry and graph cuts[J]. Computer Methods and Programs in Biomedicine, 2017, 150: 31-39. [12] Jerman T, Pernus F, Likar B, et al. Enhancement of vascular structures in 3D and 2D angiographic images[J]. IEEE Transactions on Medical Imaging, 2016, 35(9): 2107-2118. [13] Wu W, Zhou Z, Wu S, et al. Automatic liver segmentation on volumetric CT images using supervoxel-based graph cuts[J]. Computational and Mathematical Methods in Medicine, 2016, 2016: 9093721. [14] Kopp J. Efficient numerical diagonalization of hermitian 3×3 matrices[J]. International Journal of Modern Physics C, 2008, 19(3): 523-548. [15] Luu HM, Klink C, Moelker A, et al. Quantitative evaluation of noise reduction and vesselness filters for liver vessel segmentation on abdominal CTA images[J]. Physics in Medicine and Biology, 2015, 60(10): 3905-3926. [16] Zadeh LA. Fuzzy sets[M]//Fuzzy Sets, Fuzzy Logic, and Fuzzy Systems. Upper Saddle River: Prentice Hall PTR, 1996: 394-432. [17] Rosenfeld A. Fuzzy digital topology[J]. Information and Control, 1979, 40(1): 76-87. [18] Udupa JK, Samarasekera S. Fuzzy connectedness and object definition: theory, algorithms, and applications in image segmentation[J]. Graphical Models and Image Processing, 1996, 58(3): 246-261. [19] Udupa JK, Wei L, Samarasekera S, et al. Multiple sclerosis lesion quantification using fuzzy-connectedness principles[J]. IEEE Transactions on Medical Imaging, 1997, 16(5): 598-609. [20] Pednekar A, Kakadiaris IA, Kurkure U. Adaptive fuzzy connectedness-based medical image segmentation[C]//New Delhi: ICVGIP. 2002. [21] Udupa JK, Samarasekera S. Extraction of fuzzy object information in multidimensional images for quantifying ms lesions of the brain: U.S. Patent 5,812,691[P]. 1998-9-22. [22] Hsu CY, Yang CH, Wang HC. Multi-threshold level set model for image segmentation[J]. EURASIP Journal on Advances in Signal Processing, 2010, 2010(1): 950438. [23] Caselles V, Kimmel R, Sapiro G. Geodesic active contours[J]. International Journal of Computer Vision, 1997, 22(1): 61-79.
