Department of Computer Engineering, Keimyung University, Shindang-dong Dalseo-gu, Daegu, Republic of Korea.
Micron. 2011 Oct;42(7):695-705. doi: 10.1016/j.micron.2011.03.009. Epub 2011 Apr 8.
This study aims at proposing a new stained WBC (white blood cell) image segmentation method using stepwise merging rules based on mean-shift clustering and boundary removal rules with a GVF (gradient vector flow) snake. This paper proposes two different schemes for segmenting the nuclei and cytoplasm of WBCs, respectively. For nuclei segmentation, a probability map is created using a probability density function estimated from samples of WBC's nuclei and sub-images cropped to include a nucleus based on the fact that nuclei have a salient color against the background and red blood cells. Mean-shift clustering is then performed for region segmentation, and a stepwise merging scheme applied to merge particle clusters with a nucleus. Meanwhile, for cytoplasm segmentation, morphological opening is applied to a green image to boost the intensity of the granules and canny edges detected within the sub-image. The boundary edges and noise edges are then removed using removal rules, while a GVF snake is forced to deform to the cytoplasm boundary edges. When evaluated using five different types of stained WBC, the proposed algorithm produced accurate segmentation results for most WBC types.
本研究旨在提出一种新的基于均值漂移聚类和边界去除规则的分步合并规则的染色白细胞图像分割方法，结合 GVF（梯度向量流）蛇。本文分别提出了两种不同的方案来分割白细胞的细胞核和细胞质。对于细胞核分割，使用基于白细胞细胞核样本和基于包含细胞核的子图像的概率密度函数创建概率图，这是因为细胞核相对于背景和红细胞具有显著的颜色。然后进行区域分割的均值漂移聚类，并应用逐步合并方案将具有细胞核的粒子簇合并。同时，对于细胞质分割，对绿色图像进行形态学开运算，以增强子图像内检测到的颗粒和 Canny 边缘的强度。然后使用去除规则去除边界边缘和噪声边缘，同时强制 GVF 蛇变形到细胞质边界边缘。使用五种不同类型的染色白细胞进行评估时，该算法对大多数白细胞类型都产生了准确的分割结果。
