Microglial activation is a complex process that is actively under investigation. Reports have suggested both beneficial and deleterious microglial activation in neurological diseases. Attempts have been made to separate activation into subclasses with different properties. To provide further insights into the activation process, we subdivided the activated cells into 4 categories with different properties using unsupervised clustering based on their morphological features.

The first cluster represents the earliest stage of activation, with enlarged soma area compared to non-activated microglia. The second cluster represents cells with very elongated morphology, as indicated by their high moments. The third cluster includes cells with a highly branched morphology, shown by the high number of branching points in their processes. Finally, the fourth cluster represents cells in the last stages of activation, with extremely large soma and very short processes. While the activation score was not used for this clustering, it shows a gradual increase from Cluster 1 to 4, suggesting that the activation score provides a continuous metric of activation.

Having defined these clusters, we can measure how activated cells are distributed across these classes in different brain regions and animal groups. In the bottom figure, we see the distribution is similar across different brain regions, with most cells in the least activated state and the number of cells in each cluster increasing progressively in the ipsilateral hemisphere and in the piriform cortex.

In summary, identifying changes in the distribution across the subclasses, in conjunction with shifts in the activation scores, could provide further insights on disease progression or therapeutic effects that are not readily quantified simply by the density of activated cells.