Skip to main content
. 2021 Nov 8;8:750396. doi: 10.3389/fmed.2021.750396

Figure 2.

Figure 2

Our proposed VTG-Net framework. Using the CNN component (in the red box), node and edge information for generating our graph representation is extracted. A GCN model is trained using the generated graphs (in the blue box) for topological feature extraction based on the CNN features. The final output is generated by fusing the CNN output and the GCN output.