Bandekar Namrata, Wong Alexander, Clausi David, Gorbet Maud
Department of Systems Design Engineering, University of Waterloo, Ontario, Canada N2L
Annu Int Conf IEEE Eng Med Biol Soc. 2011;2011:5997-6000. doi: 10.1109/IEMBS.2011.6091482.
A novel automated cell counting technique for cell sample images used to study the side-effects of lens cleaning solutions on human corneal epithelial cells is developed. The proposed multi-step approach integrates non-maximum suppression, seeded region growing, connected component analysis, and adaptive thresholding to produce segmentation and classification results that are robust to background illumination variation and clustering of cells. The proposed algorithm is computationally efficient, and experimental results show that the average detection rate of nucleated cells is greater than 90% with the proposed technique as opposed to the state-of-the-art level set method which gives an accuracy of less than 65%.
开发了一种用于细胞样本图像的新型自动细胞计数技术，该技术用于研究镜片清洁溶液对人角膜上皮细胞的副作用。所提出的多步骤方法集成了非极大值抑制、种子区域生长、连通分量分析和自适应阈值处理，以产生对背景光照变化和细胞聚类具有鲁棒性的分割和分类结果。所提出的算法计算效率高，实验结果表明，与最先进的水平集方法相比，该技术对有核细胞的平均检测率大于90%，而水平集方法的准确率小于65%。
