Department of Bioscience Technology, Chung Yuan Christian University, Chung-Li 320314, Taiwan.
Department of Chemistry, Chung Yuan Christian University, Chung-Li 320314, Taiwan.
Int J Mol Sci. 2022 May 26;23(11):6009. doi: 10.3390/ijms23116009.
Previous methods to measure protozoan numbers mostly rely on manual counting, which suffers from high variation and poor efficiency. Although advanced counting devices are available, the specialized and usually expensive machinery precludes their prevalent utilization in the regular laboratory routine. In this study, we established the ImageJ-based workflow to quantify ciliate numbers in a high-throughput manner. We conducted number measurement using five different methods: particle analyzer method (PAM), find maxima method (FMM), trainable WEKA segmentation method (TWS), watershed segmentation method (WSM) and StarDist method (SDM), and compared their results with the data obtained from the manual counting. Among the five methods tested, all of them could yield decent results, but the deep-learning-based SDM displayed the best performance for cell counting. The optimized methods reported in this paper provide scientists with a convenient tool to perform cell counting for ecotoxicity assessment.
先前测量原生动物数量的方法大多依赖于人工计数，这种方法存在变异性大、效率低的问题。尽管有先进的计数设备，但由于其专业性和通常较高的价格，限制了它们在常规实验室常规中的广泛应用。在这项研究中，我们建立了基于 ImageJ 的工作流程，以高通量的方式定量纤毛虫数量。我们使用五种不同的方法进行数量测量：颗粒分析器法（PAM）、最大值查找法（FMM）、可训练 WEKA 分割法（TWS）、分水岭分割法（WSM）和 StarDist 法（SDM），并将它们的结果与手动计数获得的数据进行比较。在测试的五种方法中，所有方法都能得到不错的结果，但基于深度学习的 SDM 法在细胞计数方面表现出最佳性能。本文报道的优化方法为科学家们提供了一种方便的工具，用于进行细胞计数以进行生态毒性评估。
