Representative images of astrocytes detected and classified as normal (left) or hypertrophic (right) morphology using our image processing pipeline. The cell outline is shown in yellow, and the soma outline in red. GFAP is shown in red and DAPI in blue. Scale bar 10µm.
Plot showing the importance of each feature to the classification decision of the machine learning model. For each feature (e.g. soma area), this plot shows how the value of that feature (colored dot for each cell) influences the decision to classify the cell as normal or hypertrophic morphology.
The Shapley (SHAP) values shown on the graph represent the additive effects of the different features on the classification decision. Each point represents the value of the feature for a cell (color, with red indicating higher than average) and its effect on the decision, where higher SHAP values push the decision toward hypertrophic morphology.
Here are typical images of cells detected using our pipeline, separated by their classification. We can observe that the machine learning model classifies cells with very distinct morphologies into two classes.
This machine learning model is explainable and makes decisions that are consistent with the literature. For example, cells classified as hypertrophic have larger cell size and soma, higher branching, and thicker processes, than cells classified as normal morphology.