IDOL-Net: An Interactive Dual-Domain Parallel Network for CT Metal Artifact Reduction

Due to the presence of metallic implants, imaging quality computed tomography (CT) would be heavily degraded. With rapid development deep learning, several neural network models have been proposed for metal artifact reduction (MAR). Since dual-domain MAR methods can leverage hybrid information from both sinogram and image domains, they significantly improved performance compared single domain methods. However, current usually operate on domains in a specific order, which implicitly imposes certain priority prior into may ignore latent interaction between domains. To address this problem, article, we propose novel interactive parallel CT MAR, dubbed as IDOL-Net. Specifically, different existing methods, IDOL-Net is composed two modules. First, obtain high-quality guide following disentanglement that disentangles noise artifacts domain, respectively. The follow-up refinement module consists branches simultaneously sinogram, fully exploiting Extensive experiments simulated clinical data demonstrate outperforms state-of-the-art qualitative quantitative aspects.