Artifact Removal from Low-Dose Multi-Energy CT Images via GAN with Joint Loss

The emergence of multi-energy CT improves diagnostic accuracy by capturing various energy levels in one scan. However, it raises concerns about health risks from generated electromagnetic radiation. Lower radiation doses cause image artifacts. Our study introduces an image-denoising model for low-dose multi-energy CT using generative adversarial networks (GAN). We enhance the generator for multi-energy CT, incorporating a Laplacian operator-based second-order differential loss term. Our model effectively removes streak artifacts, excelling in root mean squared error (RMSE), peak signal-to-noise ratio (PSNR), and structure similarity index measure (SSIM) metrics.