Comparison of UNet3+, GAN, and DDIM for single-shot x-ray phase-contrast imaging reconstruction
Abstract
Deep learning-based single-shot x-ray phase-contrast reconstruction can suppress artifacts, reduce radiation dose and mechanical positioning requirements. This study compares UNet3+, generative adversarial network (GAN), and denoising diffusion implicit model (DDIM) for phase-contrast imaging in single absorption grating and Talbot-Lau interferometers. A paired simulation dataset was generated using a forward imaging model, and model performance was evaluated on simulated and experimental data. The results indicate that while UNet3+ achieved superior performance in peak signal-to noise ratio (PSNR) and structural similarity index (SSIM) during the training phase, the DDIM model demonstrated stronger distribution modeling capability and better generalization performance in experimental data, providing valuable references for the widespread adoption of X-ray phase-contrast imaging technology.