脑磁共振图像中改进的模糊聚类分割方法研究

潘宏亮; 郭凯文; 刘丽伟; 李阳; 侯阿临

doi:10.3788/OPE.20172513.0259

您当前的位置：

首页 >

文章列表页 >

脑磁共振图像中改进的模糊聚类分割方法研究

信息科学 | 更新时间：2020-08-13

- 脑磁共振图像中改进的模糊聚类分割方法研究
- Research on improved fuzzy clustering segmentation in magnetic resonance image for brain
- 光学精密工程 2017年25卷第10s期页码：259-265
- 作者机构：
  
  长春工业大学计算机科学与工程学院, 吉林长春 130012
- 作者简介：
- 基金信息：
  
  国家留学基金资助项目（No.201308220163）();吉林省高等教育学会2017年度高教科研课题资助项目（No.JGJX2017B13）();吉林省教育科学“十三
- DOI：10.3788/OPE.20172513.0259
  中图分类号： TP391.4
- 收稿日期：2017-06-01，
  
  修回日期：2017-08-21，
  
  纸质出版日期：2017-11-25
- 稿件说明：
移动端阅览
潘宏亮, 郭凯文, 刘丽伟等. 脑磁共振图像中改进的模糊聚类分割方法研究[J]. 光学精密工程, 2017,25(10s): 259-265

PAN Hong-liang, GUO Kai-wen, LIU Li-wei etc. Research on improved fuzzy clustering segmentation in magnetic resonance image for brain[J]. Editorial Office of Optics and Precision Engineering, 2017,25(10s): 259-265
潘宏亮, 郭凯文, 刘丽伟等. 脑磁共振图像中改进的模糊聚类分割方法研究[J]. 光学精密工程, 2017,25(10s): 259-265 DOI： 10.3788/OPE.20172513.0259.

PAN Hong-liang, GUO Kai-wen, LIU Li-wei etc. Research on improved fuzzy clustering segmentation in magnetic resonance image for brain[J]. Editorial Office of Optics and Precision Engineering, 2017,25(10s): 259-265 DOI： 10.3788/OPE.20172513.0259.

摘要

模糊C均值聚类（Fuzzy C Mean clustering，FCM）算法因其聚类中心初始化敏感和陷入局部最优的缺点不能对非线性可分离数据进行聚类，而初始聚类中心的选择是随机选取，导致图像分割结果较差。为了克服这一缺点，提出了高斯核模糊C均值聚类算法。首先为保证传统FCM算法能够快速收敛到准确稳定的聚类中心，提出一种模拟退火遗传算法的模糊C均值聚类算法。该算法利用模拟退火较强的局部搜索能力和遗传算法较强的全局搜索能力，可以有效地选取初始聚类中心，提高收敛速度。为增强改进算法对聚类结果不理想的问题，借助高斯核函数进行非线性变换映射至高维空间，转换为高维空间的线性可分问题，改善脑MRI图像分割效果。在不同类型噪声和偏移场的影响下，本文算法分割时间比其它算法节约1~3倍，概率兰德指数（Probabilistic Rand Index，PRI）至少增加0.01，可以发现本文算法在临床图像分割结果中，分割时间和脑白质与灰质分割精度都有明显优势。对临床脑MRI的图像分割，本文算法比传统FCM算法在速度和精度上具有更好的鲁棒性与有效性。

Abstract

The main defect for Fuzzy C mean clustering (FCM) was initialization sensitivity and trapping into the local optimum of clustering center

and nonlinear separable data could not be provided with clustering through this algorithm. Selection for initial clustering center was random selection

which causes relatively bad image segmentation results. In order to conquerthis defect

Gaussian kernel Fuzzy C mean clustering was proposed. Fuzzy C mean clustering algorithm for stimulated annealing genetic algorithm was proposed to guarantee traditional algorithm can rapidly converge to accurate and stable clustering center. Relatively strong local search ability for simulated annealing and relatively strong global search ability for genetic algorithm were utilized to effectively select initial clustering center and improved convergence speed. In order to strengthen and improve non-ideal problem of improved algorithm to clustering result

nonlinear transformation was conducted with the help of Gaussian kernel function to be mapped into higher dimensional space and be conversed into linear separable problem of higher dimensional space

which could improve brain MRI image segmentation effect. Under the effect of noises and biased field with different types

segmentation time for algorithm in the thesis saved 1-3 times than other algorithms

and it was at least 0.01 larger than Probabilistic Rand Index (PRI). It can be found that time

white matter and grey matter segmentation accuracies of algorithm in the thesis both have obvious advantages in clinical image segmentation results. For brain MRI image segmentation

the algorithm has better robustness and effectiveness in speed and accuracy compared with traditional FCM algorithm.

关键词

Keywords

references

RULANINGTYAS R, SUKSMONO A B, MENGKO T, et al.. Multi patch approach in K-means clustering method for color image segmentation in pulmonary tuberculosis identification[C]. International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering. IEEE, 2015:75-78.

王和春,宫迅凯. 用模糊聚类Fuzzy C-Means算法实现图像分割[J]. 光学机械, 1991(4):65-70. WANG H CH, GONG X K. By using the fuzzy clustering algorithm for image segmentation Fuzzy C-Means[J]. Optical Machinery, 1991(4):65-70.(in Chinese)

BEZDEK J C, EHRLICH R, FULL W. FCM:The fuzzy c-means clustering algorithm[J]. Computers & Geosciences, 1984, 10(2):191-203.

张怀柱,向长波,宋建中,等. 改进的遗传算法在实时图像分割中的应用[J]. 光学精密工程, 2008,(2):333-337. ZHANG H ZH, XIANG CH B, SONG J ZH, Qiao S.Application of improved genetic algorithm in real-time image segmentation[J]. Opt. Precision Eng., 2008, (2):333-337.(in Chinese)

KAVITHA A R, CHELLAMUTHU C. Brain tumour segmentation from MRI image using genetic algorithm with fuzzy initialisation and seeded modified region growing (GFSMRG) method[J]. Imaging Science Journal, 2016, 64(5):1-13.

古良玲,王玉菡. 基于模拟退火遗传算法的图像增强[J]. 激光杂志, 2015(2):19-22. GU L L, WANG Y N. Image enhancement based on simulated annealing genetic algorithm[J]. Journal of Lasers, 2015(2):19-22.(in Chinese)

唐苦, 王昕, 王振雷,等. 一种基于高斯核化有效性指标的自适应优选聚类数的FKCM[J]. 计算机与应用学,2012,29(10):1199-1203. TANG K, WANG X, WANG ZH L, et al.. An adaptive optimization clustering number based on Gauss's nuclear validity index FKCM[J]. Computer and Applied Chemistry, 2012, 29(10):1199-1203.(in Chinese)

http://brainweb.bic.mni.mcgill.ca/brainweb/

AHMED E, WANG C, JIA F, et al.. Segmentation of brain tissues from magnetic resonance images using adaptively regularized kernel-based fuzzyC-means clustering[J]. Computational & Mathematical Methods in Medicine, 2015, 2015:485495.

SINGH P, BHADAURIA H S, SINGH A. Automatic brain MRI image segmentation using FCM and LSM[C]. International Conference on Reliability, INFOCOM Technologies and Optimization, IEEE, 2015:1-6.

http://www.loni.usc.edu/

浏览量

465

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

融合条形注意力机制的电芯极片计算机断层图像分割

增强型金枪鱼群优化指数熵的砂粒显微图像分割

语义流引导采样结合注意力机制的脑肿瘤图像分割

混沌SSA优化多重熵阈值的骨料图像自动分割

基于高分辨率复合网络的皮肤病变分割