The animation at the top shows the overall process for the regional analysis of microglial morphology on IHC sections. Briefly, after we stain the tissue sections for Iba-1 using an automated immunohistochemistry stainer, we digitize the slides using a whole slide scanner. After scanning, we get a high-resolution digital image with sub-micrometer spatial resolution.  

We then apply the approach outlined in the schematic at the bottom of the page for segmentation and classification of the microglia over the entire tissue section. The IHC stain is first segmented, and then the cells are detected and classified using computer vision algorithms. The microglia are skeletonized and their processes are separated from the soma. This approach enables the extraction of a high number of morphological features, such as the cell size, the soma size, the number of branching points in the skeleton, etc. These features are then used by a machine learning model that assigns an activation score to each cell and then classifies that cell as activated or not. 

We then use deep-learning artificial intelligence (AI) algorithms to label the neuroanatomy on tissue sections. By analyzing an entire brain region, such as the piriform cortex, hippocampus, or olfactory bulb, we can improve the reliability of the measurements compared with simple analysis of random high-magnification snapshots..

Finally, the results are then aggregated by region-of-interest, subject, and group. In addition to the stain density, our platform can output the total cell density, the activated cell density, and the histogram of cell properties, such as the distribution of activation scores. We will discuss that further in this presentation.

This entire image processing pipeline has been implemented in a scalable manner, such that large data sets can be rapidly analyzed. We are able to analyze microglial activation from all tissue sections from all mice in efficacy studies of putative disease-modifying therapeutic agents.