Skip to main content
. 2021 May;9(9):819. doi: 10.21037/atm-20-3345

Figure 7.

Figure 7

The progress when training the neural network converting projection data and attenuation map to compressed image. The RMSE reduced very quickly and saturated when the training reached around 20 epochs. Data shown in this plot are the 2,000 cases for the validation purpose.