Dual Domain Reconstruction

Dual-domain reconstruction leverages information from both the image and its transform domain (e.g., k-space in MRI, Radon domain in CT) to improve the accuracy and efficiency of reconstructing images from incomplete data. Current research focuses on developing deep learning models, often employing hybrid architectures like combinations of convolutional neural networks (CNNs) and transformers, to effectively fuse information across domains and handle varying data quality. These advancements are significantly impacting medical imaging, enabling faster scans with reduced radiation exposure and improved image quality in applications such as MRI and CT, and also showing promise in other areas like face forgery detection and 3D surface reconstruction.