Marcuzzo Monica, Quelhas Pedro, Campilho Ana, Mendonça Ana Maria, Campilho Aurélio
INEB - Instituto de Engenharia Biomédica, Divisão de Sinal e Imagem, Campus FEUP, Portugal.
Comput Biol Med. 2009 Sep;39(9):785-93. doi: 10.1016/j.compbiomed.2009.06.008. Epub 2009 Jul 14.
To obtain development information of individual plant cells, it is necessary to perform in vivo imaging of the specimen under study, through time-lapse confocal microscopy. Automation of cell detection/marking process is important to provide research tools in order to ease the search for special events, such as cell division. In this paper we discuss an automatic cell detection approach for Arabidopsis thaliana based on segmentation, which selects the best cell candidates from a starting watershed-based image segmentation and improves the result by merging adjacent regions. The selection of individual cells is obtained using a support vector machine (SVM) classifier, based on a cell descriptor constructed from the shape and edge strength of the cells' contour. In addition we proposed a novel cell merging criterion based on edge strength along the line that connects adjacent cells' centroids, which is a valuable tool in the reduction of cell over-segmentation. The result is largely pruned of badly segmented and over-segmented cells, thus facilitating the study of cells. When comparing the results after merging with the basic watershed segmentation, we obtain 1.5% better coverage (increase in F-measure) and up to 27% better precision in correct cell segmentation.
为了获取单个植物细胞的发育信息，有必要通过延时共聚焦显微镜对研究中的样本进行活体成像。细胞检测/标记过程的自动化对于提供研究工具很重要，以便于寻找诸如细胞分裂等特殊事件。在本文中，我们讨论了一种基于分割的拟南芥自动细胞检测方法，该方法从基于分水岭的初始图像分割中选择最佳细胞候选对象，并通过合并相邻区域来改进结果。使用支持向量机（SVM）分类器，基于从细胞轮廓的形状和边缘强度构建的细胞描述符来选择单个细胞。此外，我们提出了一种基于连接相邻细胞质心的线的边缘强度的新型细胞合并标准，这是减少细胞过度分割的一个有价值的工具。结果在很大程度上剔除了分割不良和过度分割的细胞，从而便于细胞研究。将合并后的结果与基本分水岭分割进行比较时，我们在正确细胞分割方面获得了1.5%的更好覆盖率（F值增加）和高达27%的更好精度。
