DDHANET: Dual-Domain Hybrid Attention-Guided Network for Ct Scatter Correction

Scattered x-ray photons generated by the interaction between objects and x-rays will produce artifacts in X-ray computed tomography (CT) images and detrimentally affect the diagnosis. The success of deep learning has inspired us to reduce scatter artifacts in CT images. In this paper, we propose a novel dual-domain network combined with hybrid attention modules. We train the network to reduce artifacts in the projection and image domains. Hybrid attention modules help the network focus on important information and suppress less relevant information. In addition, we describe a joint loss function to preserve the image details. Compared with other state-of-the-art methods, our proposed method shows that it can improve peak signal-to-noise and structural similarity.