面向医学图像分割的蚁群密度峰值聚类  

Ant Colony Algorithm and Density Peaks Clustering for Medical Image Segmentation

作　　者：朱红 何瀚志 方谦昊 代岳 姜代红[3] Zhu Hong;He Hanzhi;Fang Qianhao;Dai Yue;Jiang Daihong(School of Medical Information,Xuzhou Medical University,Xuzhou 221004,China;Radiology Department,The Affiliated Hospital of Xuzhou Medical University,Xuzhou 221002,China;School of Information and Electrical Engineering,Xuzhou Institute of Technology,Xuzhou 221008,China)

机构地区：[1]徐州医科大学医学信息学院,江苏徐州221004 [2]徐州医科大学附属医院影像科,江苏徐州221002 [3]徐州工程学院信电工程学院,江苏徐州221008

出　　处：《南京师大学报（自然科学版）》2019年第2期1-8,共8页Journal of Nanjing Normal University(Natural Science Edition)

基　　金：国家自然科学基金项目(61672522);江苏省高等学校自然科学研究重大项目(18KJA520012);徐州市科技计划项目(KC16SQ78)

摘　　要：在医学图像分割研究中,针对密度峰值聚类算法(density peaks clustering algorithm,DPC),依靠先验知识给定截断距离dc且人工选择聚类中心点具有主观随意性等缺陷,提出了一种结合蚁群算法选取密度峰值聚类最优参数的医学图像分割方法.该算法首先利用蚁群算法全局性和鲁棒性的优点,使用图像熵计算信息素来指导蚁群的搜索路径;再使用变量量化表示聚类中心个数,蚁群通过迭代选择最优截断距离dc和聚类中心,实现了DPC算法的自适应分割并得到了较好的分割效果.仿真实验分析证明了算法的有效性和实用性.In the medical image segmentation research,a medical image segmentation method based on ant colony algorithm for selecting the optimal parameters of density peaks clustering(DPC)was proposed. For some defects in DPC algorithm,such as cut-off distance d c was given by DPC algorithm relied on prior knowledge,subjective randomness in cluster centers was selected by manual work. First,the algorithm took advantages of the overall robustness of the ant colony algorithm,used image entropy to calculate pheromone to guide the search path of ant colony. Then it was quantified the number of cluster centers by using variable quantification instead,and ant colony selected the optimal truncation distance d c and cluster centers by iteration,and realized the adaptive segmentation of DPC algorithm and obtained better results. Simulation experiments proved the effectiveness and practicability of this algorithm.

关 键 词：医学图像分割 密度峰值 聚类中心 蚁群算法 信息素 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]