Abstract Counting cells is a cornerstone of tracking disease progression in neuroscience. A common approach for this process is having trained researchers individually select and count cells within an image, which is not only difficult to standardize but also very time-consuming. We present an automatic cell counting methodology which leverages the Trainable WEKA Segmentation (TWS) plugin available in the FIJI distribution of ImageJ (Schindelin 2012, Schneider 2012). TWS provides access to machine learning tools for object segmentation with an accessible graphical user interface. However, to analyze whole datasets requires further automation through the use of macros which we constructed within ImageJ maintain this high level of accessibility. We compared our automatic counts to a published, hand counted dataset of Immunofluorescence stained mouse cortex microglia (Singh et al. 2020). For this analysis we used a wild-type (WT) control and a transgenic mouse model for HIV-induced brain injury (HIVgp120). HIVgp120 mice showed behavioral deficits, and increase in microglia numbers presumably caused the gp120 protein. 10X images were collected in the sagittal plane from brain slices of three 11–14 month-old male mice for each genotype. The auto counting strategy with TWS replicated the increase in mean microglial density in images from the HIVgp120 mice (Manual p =0.0001; Auto p=0.0002). The auto count also correlates significantly with the manual counts in both genotypes (WT p= 2.297 e−07, adj. R2= 0.65; HIVgp120 p=2.591 e−04 adj. R2=0.423). The auto counting strategy took less than a fifth of the time for the expert manual count, allowing for the efficient exploration of large sets of image data.

An Accessible Automatic Cell Counting Methodology Using Trainable Weka Segmentation in ImageJ Validated on Murine Microglia