Fig. 4: Overlapping of cell clusters in the ganglionic eminence. | Nature Communications

Fig. 4: Overlapping of cell clusters in the ganglionic eminence.

From: Unraveling microglial spatial organization in the developing human brain with DeepCellMap, a deep learning approach coupled with spatial statistics

Fig. 4

A DBSCAN application to the (x,y) coordinates of the cells' center of mass. B Graph of MinSample versus ϵ used to select the value of the radius ϵ that maximizes the number of clusters. C Schematic representation of the procedure to select stable clusters by removing 10% of the cell located on the edges of the convex hull while the area of the new convex hull remains  > 60% of the initial surface. D Clustering of the five morphological states using DBSCAN algorithms, generating convex hulls around each cluster. Color code is the same as microglial morphological types: proliferative (pink), amoeboid (light blue), aggregated (green), phagocytic (yellow), and ramified (dark blue). E Scheme of isolated vs clustered cells. F Percentage of isolated vs clustered cells for the five populations at 17 pcw in the ganglionic eminence. G Time-dependent clustering percentage at 17,19, and 20 pcw. H Scheme of intersecting clusters from two different populations (type 1 vs type 2) and associated metric. I Cluster mixing rate matrix between the 5 microglial morphological states in the ganglionic eminence at 17 pcw. J % of four microglia types vs ramified cluster across time at 17,19 and 20 pcw. Source data are provided as a Source Data file.

Back to article page