Quantification of microglial activation through morphometric analysis has long been a staple of the neuroimmunologist’s toolkit. Microglial morphological phenomics can be conducted through either manual classification or constructing a digital skeleton and extracting morphometric data from it. Multiple open-access and paid software packages are available to generate these skeletons via semi-automated and/or fully automated methods with varying degrees of accuracy. Despite advancements in methods to generate morphometrics (quantitative measures of cellular morphology), there has been limited development of tools to analyze the datasets they generate, in particular those containing parameters from tens of thousands of cells analyzed by fully automated pipelines. In this review, we compare and critique the approaches using cluster analysis and machine learning driven predictive algorithms that have been developed to tackle these large datasets, and propose improvements for these methods. In particular, we highlight the need for a commitment to open science from groups developing these classifiers. Furthermore, we call attention to a need for communication between those with a strong software engineering/computer science background and neuroimmunologists to produce effective analytical tools with simplified operability if we are to see their wide-spread adoption by the glia biology community.

中文翻译：

---

小胶质细胞形态分析：选项太多，一致性太差

通过形态测量分析对小胶质细胞激活进行量化长期以来一直是神经免疫学家工具包的主要内容。小胶质细胞形态表型组学可以通过手动分类或构建数字骨架并从中提取形态测量数据来进行。多个开放获取和付费软件包可通过半自动和/或全自动方法以不同程度的准确度生成这些骨架。尽管生成形态测量（细胞形态的定量测量）的方法取得了进步，但用于分析它们生成的数据集的工具的开发仍然有限，特别是那些包含由全自动管道分析的数万个细胞的参数的工具。在这篇综述中，我们比较和批评了使用聚类分析和机器学习驱动的预测算法来处理这些大型数据集的方法，并提出了对这些方法的改进。我们特别强调需要做出承诺开放科学来自开发这些分类器的小组。此外，如果我们希望看到神经免疫学家广泛采用具有强大软件工程/计算机科学背景的人员和神经免疫学家之间的沟通，以产生具有简化可操作性的有效分析工具，我们呼吁人们注意这些工具。