A geometry-informed deep learning framework for ultra-sparse 3D tomographic image reconstruction. (September 2022)

• Record Type: Journal Article
• Authors: Shen, Liyue; Zhao, Wei; Capaldi, Dante; Pauly, John; Xing, Lei
• Is Part Of: Computers in biology and medicine. Volume 148(2022)
• Publication Date: 2022-09
• Subjects: Geometry-informed deep learning -- Sparse-view 3D image reconstruction -- Deep learning -- Image reconstruction; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2022.105710
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: Deep learning affords enormous opportunities to augment the armamentarium of biomedical imaging. However, the pure data-driven nature of deep learning models may limit the model generalizability and application scope. Here we establish a geometry-informed deep learning framework for ultra-sparse 3D tomographic image reconstruction. We introduce a novel mechanism for integrating geometric priors of the imaging system. We demonstrate that the seamless inclusion of known priors is essential to enhance the performance of 3D volumetric computed tomography imaging with ultra-sparse sampling. The study opens new avenues for data-driven biomedical imaging and promises to provide substantially improved imaging tools for various clinical imaging and image-guided interventions. Highlights: We propose a geometry-informed deep learning framework for ultra-sparse 3D tomographic image reconstruction (GIIR) by integrating geometric priors of the imaging system to bridge 2D and 3D image space. We introduce a new 2D projection generation network to synthesize novel-view X-ray projections via representation disentanglement, which further contributes to 3D tomographic image reconstruction. We validate the proposed method using large-scale 3D lung images and demonstrate that the known priors are essential to enhance the performance of 3D tomographic imaging with ultra-sparse sampling.