Unpaired Low-Dose CT Denoising using Conditional GAN with Structural Loss
Zhao Yang, Chenggeng Yan, Hu Chen · 2021 International Conference on Wireless Communications and Smart Grid (ICWCSG) · 2021
Low-dose CT has been a popular diagnostic imaging for its high availability and less radiation than normal-dose CT. Reducing the noise and reconstructing a noise-free CT image is a hotspot for researchers. The existing methods cannot deal with unpaired data well yet while the paired data is difficult to obtain. With unpaired data, we proposed an effective method using conditional generative adversarial network with Structural Loss to solve this problem. Our method is simpler and more efficient on computation than other unpaired methods and our experiments showed that it is comparable to other deep learning methods on denoising performance.