K-Means Clustering

To analyze the features of the giant PyNs soma five data blocks in the ferret motor cortex at a voxel size of 300 μm × 300 μm × 300 μm were randomly selected, and all stained cells in these blocks were manually segmented using the segmentation editor module of Amira v6.1.1 software (FEI, Villebon sur Yvette, France); the volume, mean gray value, and longest radius of each cell were then calculated.

To analyze differences in cell volume between giant PyNs and other PyNs the two cell types were manually identified according to previously defined criteria (White et al., 1997; Rivara et al., 2003). To further analyze giant PyNs K-means clustering, an unsupervised machine learning technique, was performed for all cells with cell volume, mean gray value, and longest radius as the three principal components (Xu and Tian, 2015; Arora et al., 2016). The cluster centroids representing the center of data points of giant PyNs and other cells were iteratively updated until objects in the same cluster showed high similarity while those in different clusters showed lower similarity.