Division of Medical Sciences, University of Victoria, Victoria, BC, Canada.
Division of Medical Sciences, University of Victoria, Victoria, BC, Canada; Axe Neurosciences, Centre de recherche du CHU de Québec-Université Laval, Québec, Qc, Canada; Department of Molecular Medicine, Université de Laval, Québec City, Qc, Canada.
Micron. 2022 Oct;161:103334. doi: 10.1016/j.micron.2022.103334. Epub 2022 Aug 2.
Microglia, the immune resident cells of the central nervous system (CNS), are now recognized as performing crucial roles for maintaining homeostasis and determining the outcomes of various pathological challenges across life. While brightfield microscopy is a powerful and established tool to study microglia-mediated mechanisms underlying neurological diseases, microglial density and distribution are some of the most frequently investigated parameters. Their quantitative assessment provides relevant clues regarding dynamic densitometric changes in the microglial population across various CNS regions. Investigators often rely on a manual identification and analysis of these cells within key regions of interest, which can be time-consuming and introduce an experimenter bias. Automation of this process, which has been gaining popularity in recent years, represents a potential solution to minimize both experimenter's bias and time investment, thus increasing the efficacy of the experiment and uniformity of the collected data. We aimed to compare manual versus automatic analysis methods to determine whether an automatic analysis is efficient and accurate enough to replace a manual analysis in both homeostatic and pathological contexts (i.e., adult healthy and lipopolysaccharide-challenged adolescent male mice, respectively). To do so, we used a script that runs on the ImageJ software to perform microglial density analysis by automatic detection of microglial cells from brightfield microscopy images. The main core of the macro script consists in an automatic cell selection step using a threshold followed by a spatial analysis for each selected cell. The resulting data were then compared with the values obtained using a well-established manual method. Overall, the evaluation of the established automatic densitometry method with manual density and distribution analysis revealed similar results for the density and nearest neighbor distance in healthy adult mice, as well as density and distribution in lipopolysaccharide-challenged adolescent mice. Applying machine learning to the automatic process could further improve the accuracy and robustness of the method.
小胶质细胞是中枢神经系统（CNS）的免疫常驻细胞，现在被认为在维持内稳态和决定各种病理挑战的结果方面发挥着关键作用。虽然明场显微镜是研究神经疾病中小胶质细胞介导机制的强大而成熟的工具，但小胶质细胞密度和分布是最常研究的参数之一。它们的定量评估提供了关于各种中枢神经系统区域中小胶质细胞群体动态密度变化的相关线索。研究人员通常依赖于在关键感兴趣区域内手动识别和分析这些细胞，这可能既耗时又引入实验者偏见。近年来，这种过程的自动化越来越受欢迎，它代表了一种潜在的解决方案，可以最大限度地减少实验者偏见和时间投入，从而提高实验的效率和收集数据的一致性。我们旨在比较手动与自动分析方法，以确定自动分析是否在稳态和病理情况下（即成年健康和脂多糖挑战的青春期雄性小鼠）都足够高效和准确，可以替代手动分析。为此，我们使用了一个在 ImageJ 软件上运行的脚本，通过自动检测明场显微镜图像中的小胶质细胞来进行小胶质细胞密度分析。宏脚本的主要核心是使用阈值进行自动细胞选择步骤，然后对每个选定的细胞进行空间分析。然后将得到的数据与使用成熟的手动方法获得的值进行比较。总的来说，使用手动密度和分布分析对建立的自动密度计量方法的评估显示，在健康成年小鼠中，密度和最近邻距离以及脂多糖挑战的青春期小鼠中的密度和分布的结果相似。将机器学习应用于自动过程可以进一步提高方法的准确性和稳健性。
