A comparison of machine learning approaches for the quantification of microglial cells in the brain of mice, rats and non-human primates

• DOI: 10.1371/journal.pone.0284480
• Publication Date: 2023-05-01T17:24:27Z
• Authors: Danish M. Anwer; Francesco Gubinelli; Yunus A. Kurt; Livija Sarauskyte; Febe Jacobs; Chiara Venuti; Ivette M. Sandoval; Yiyi Yang; Jennifer Stancati; Martina Mazzocchi; Edoardo Brandi; Gerard O’Keeffe; Kathy Steece-Collier; Jia-Yi Li; Tomas Deierborg; Fredric P. Manfre...; Marcus Davidsson; Andreas Heuer
• Keywords: Robustness; Human brain

Microglial cells are brain-specific macrophages that swiftly react to disruptive events in the brain. activation leads specific modifications, including proliferation, morphological changes, migration site of insult, and changes gene expression profiles. A change inflammatory status has been linked many neurodegenerative diseases such as Parkinson’s disease Alzheimer’s disease. For this reason, investigation quantification microglial is essential for better understanding their role progression well evaluating cytocompatibility novel therapeutic approaches conditions. In following study we implemented a machine learning-based approach fast automatized cells; tool was compared with manual (ground truth), alternative free-ware threshold-based ImageJ Ilastik. We first trained algorithms on brain tissue obtained from rats non-human primate immunohistochemically labelled microglia. Subsequently validated accuracy preclinical rodent model demonstrated robustness mice, images provided by three collaborating laboratories. Our results indicate learning can detect quantify all mammalian species precise manner, equipotent one observed counting. Using tool, were able small between hemispheres, suggesting power reliability algorithm. Such will be very useful response development, compatible therapeutics targeting As network weights training data made available, together our step-by-step user guide, anticipate laboratories implement research.