Because we use the denoising model of MIRNet and did not train on CT images, the effect is general. If you train enough on the CT model, I think the effect should be satisfactory. The result is

The input image is（added gaussian noise）

Dataset：https://pan.baidu.com/s/1UA843ZUphDGidSuwXMebfQ access：mao9

The code is updated on https://github.com/swz30/MIRNet