Bunyak Filiz, Palaniappan Kannappan, Chagin Vadim, Cardoso M
Department of Computer Science, University of Missouri-Columbia, Columbia MO 65211-2060, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:1424-8. doi: 10.1109/IEMBS.2009.5334168.
Fluorescently tagged proteins such as GFP-PCNA produce rich dynamically varying textural patterns of foci distributed in the nucleus. This enables the behavioral study of sub-cellular structures during different phases of the cell cycle. The varying punctuate patterns of fluorescence, drastic changes in SNR, shape and position during mitosis and abundance of touching cells, however, require more sophisticated algorithms for reliable automatic cell segmentation and lineage analysis. Since the cell nuclei are non-uniform in appearance, a distribution-based modeling of foreground classes is essential. The recently proposed graph partitioning active contours (GPAC) algorithm supports region descriptors and flexible distance metrics. We extend GPAC for fluorescence-based cell segmentation using regional density functions and dramatically improve its efficiency for segmentation from O(N(4)) to O(N(2)), for an image with N(2) pixels, making it practical and scalable for high throughput microscopy imaging studies.
诸如绿色荧光蛋白标记的增殖细胞核抗原（GFP-PCNA）等荧光标记蛋白会在细胞核中产生丰富的、动态变化的焦点纹理模式。这使得在细胞周期的不同阶段对亚细胞结构进行行为研究成为可能。然而，荧光的点状模式变化、有丝分裂期间信噪比、形状和位置的剧烈变化以及接触细胞的数量众多，都需要更复杂的算法来进行可靠的自动细胞分割和谱系分析。由于细胞核外观不均匀，基于分布的前景类别建模至关重要。最近提出的图分割活动轮廓（GPAC）算法支持区域描述符和灵活的距离度量。我们使用区域密度函数扩展了GPAC用于基于荧光的细胞分割，并将其分割效率从O(N(4))显著提高到O(N(2))，对于具有N(2)个像素的图像，使其适用于高通量显微镜成像研究且具有可扩展性。
