Skip to main content
. 2021 Dec 2;11:23285. doi: 10.1038/s41598-021-02683-4

Figure 1.

Figure 1

Overview of the HETEROID framework. VSMC spheroid images (A) were segmented by VGG16-U-Net semantic segmentation (B), and feature vectors were extracted for additional clustering analyses. A two-level clustering framework (C) was used to analyze the presence of morphological subpopulations of VSMC spheroids. Initial clustering analysis was focused on the roundness of the spheroids. Subsequent clustering analysis was focused on disrupted spheroids.