Abstract
Microglia are dynamic central nervous system cells crucial for maintaining homeostasis and responding to neuroinflammation, as evidenced by their varied morphologies. Existing morphology analysis often fails to detect subtle variations within the full spectrum of microglial morphologies due to their reliance on predefined categories. Here, we present MorphoGlia, an interactive, user-friendly pipeline that objectively characterizes microglial morphologies. MorphoGlia employs a machine learning ensemble to select relevant morphological features of microglia cells, perform dimensionality reduction, cluster these features, and subsequently map the clustered cells back onto the tissue, providing a spatial context for the identified microglial morphologies. We applied this pipeline to compare the responses between saline solution (SS) and scopolamine (SCOP) groups in a SCOP-induced mouse model of Alzheimer's disease, with a specific focus on the hippocampal subregions CA1 and Hilus. Next, we assessed microglial morphologies across four groups: SS-CA1, SCOP-CA1, SS-Hilus, and SCOP-Hilus. The results demonstrated that MorphoGlia effectively differentiated between SS and SCOP-treated groups, identifying distinct clusters of microglial morphologies commonly associated with pro-inflammatory states in the SCOP groups. Additionally, MorphoGlia enabled spatial mapping of these clusters, identifying the most affected hippocampal layers. This study highlights MorphoGlia's capability to provide unbiased analysis and clustering of microglial morphological states, making it a valuable tool for exploring microglial heterogeneity and its implications for central nervous system pathologies.