Med Biol Eng Comput. 2013 Dec;51(12):1325-38. doi: 10.1007/s11517-013-1105-y.
Since the introduction of the automated karyotyping systems, segmentation and classification of touching and overlapping chromosomes in the metaphase images are major challenges. The earlier reported techniques for disentangling the chromosome overlaps have limited success and use only color information in case of multispectral imaging. Most of them are restricted to separation of single overlap of two chromosomes. This paper introduces a novel algorithm to extricate overlapping chromosomes in a metaphase image. The proposed technique uses Delaunay triangulation to automatically identify the number of overlaps in a cluster followed by the detection of the appropriate cut-points. The banding information on the overlapped region further resolves the set of overlapping chromosomes with the identified cut-points. The proposed algorithm has been tested with four data sets of 60 overlapping cases, obtained from publically available databases and private genetic labs. The experimental results provide an overall accuracy of 75–100 % for resolving the cluster of 1–6 overlaps.
自从自动化核型分析系统问世以来，如何对中期图像中相互接触和重叠的染色体进行分割和分类一直是个难题。早期报道的用于解开染色体重叠的技术仅取得了有限的成功，而且仅在多光谱成象的情况下使用颜色信息。这些技术大多仅限于分离两个染色体的单个重叠。本文介绍了一种新颖的算法，可以从中期图象中提取相互重叠的染色体。该技术使用 Delaunay 三角剖分自动识别群集中的重叠数量，然后检测适当的分割点。重叠区域上的带型信息进一步解决了具有识别出的分割点的重叠染色体集。该算法已在四个重叠案例数据集（分别来自公开数据库和私人基因实验室）上进行了测试，重叠案例数为 60。实验结果表明，该算法用于解析 1-6 个重叠的聚类的总体准确率为 75-100%。
