一种新的基于细胞局部图特征和运动特征的细胞跟踪算法  

A NOVEL CELL TRACKING ALGORITHM BASED ON CELL LOCAL GRAPH FEATURE AND MOTION FEATURE

作　　者：朱辰[1] 周庆利[1] 何剑虎[1] 

机构地区：[1]浙江大学医学院附属妇产科医院,浙江杭州310006

出　　处：《计算机应用与软件》2016年第7期181-185,251,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61173096)

摘　　要：在跟踪显微序列图像中的细胞时,由于存在细胞分裂、粘连运动不均匀等问题,影响细胞分割和跟踪的准确性。为了提高细胞跟踪的准确率,提出一种新的跟踪算法,用于分析显微序列图像中细胞的形态变化和动态变化。该算法对移动距离不大的细胞采用由细胞位移、面积和形变构成的细胞区域特征来实现该类细胞的跟踪,然后再利用已成功实现跟踪的细胞,构造由细胞之间拓扑关系组成的局部图,最终使用基于局部图特征和细胞区域特征构成的混合跟踪算法实现活跃细胞的跟踪,并对跟踪过程中的各种情况进行分析处理。用两个视频序列验证该方法的有效性,实验结果表明,该算法能够提高细胞跟踪的准确率分别为10%和6%。The presence of cell division,adhesion and uneven movement affect the accuracy of cell segmentation and tracking when the cell in microscopic sequence images is tracked. In order to improve the cell tracking accuracy,this paper proposes a novel tracking algorithm used for analysing the morphological change and dynamical change of cell in microscopic sequence image. For those cells with little moving distance,this algorithm uses the cell region features consisted of cell displacement,area and deformation to achieve the track on them,and then makes use of these successfully tracked cells to construct the local graph composed of the topological relations between cells,and finally uses the hybrid tracking algorithm,which is constructed based on local graph features and cell region features,to achieve the track on the active cells,and analyses and processes various situations appeared during cell tracking process. Two video sequences are used to verify the effectiveness of this method. Experimental results show that the proposed algorithm can improve 10% and 6% in cell tracking accuracy respectively.

关 键 词：细胞分割 细胞跟踪 聚集细胞 局部图 时间序列图像 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]