Skip to main content
. 2021 Oct 4;2021:4196241. doi: 10.1155/2021/4196241

Figure 7.

Figure 7

The superlabel-guided convolutional neural network (SG-CNN) structure. A raw CT image with two labels without any annotations is fed into the model. The input image is cropped under the guide of the heatmap generated by the first convolutional layer of the superlabel subnetwork and then inserted into the other network branch. The network output provides two labels [71].