Skip to main content
. 2021 Apr 16;11:8381. doi: 10.1038/s41598-021-87737-3

Figure 4.

Figure 4

Grad-CAM results of the cannulation difficulty prediction model for accurately predicted examples. The heatmap visualizations are the outputs of Grad-CAM for the written GT label. They show where the model focused attention to estimate the cannulation difficulty. The lighter the color, the stronger is the attention required. The results show that the model sees distinct features of each label.