基于隐马尔可夫随机场的细胞分割方法

Cell segmentation method based on hidden Markov random field

为了提高细胞聚合、粘连区域的分割准确性,本文提出一种基于空间聚类和隐马尔可夫随机场的两级分割算法。以像素点颜色特征为依据,在Lab色彩空间中采用k-means++聚类方法得到初始化标签集;通过HMRF构建细胞图像的空间表达模型,充分利用空间约束关系,减少孤立点影响,平滑分割区域;采用期望最大值算法优化模型参数,利用标记场和观测场的相互作用,通过迭代算法不断调整标签集合,迭代直至收敛得到全局最优值。对来自于骨髓涂片的61幅细胞图像中的780个6类细胞的实验表明,本文算法提高了分割的准确率(不小于95%),便于进一步提取细胞病理特征,实现检测识别。

A two-level segmentation algorithm based on spatial clustering and hidden Markov random field(HMRF)is proposed to improve the segmentation accuracy of cell aggregation and adhesion region.First,based on color feature of pixels in the Lab color space,k-means++clustering method is used to obtain an initialization tag set.Second,the spatial expression model of the cell image is constructed by HMRF,which fully employs the spatial constraint relation to reduce the influence of isolated points and smooth the segmentation area.Finally,the model parameters are optimized by using the expectation maximization algorithm.Through the interaction between the marker and observation fields,the label set is adjusted by the iterative algorithm.Experimental results of six kinds of 780 cells from bone marrow smears of 61 cell types show that the proposed algorithm improves the accuracy of segmentation by≥95%.Furthermore,the algorithm is convenient for further extraction,detection,and recognition of cell pathology characteristics.