2019/06/13 by Valery Vishnevskiy, Richard Rau, Vishnevskiy, Valery +3
Engineering · Medicine · #Advanced X-ray and CT Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Photoacoustic and Ultrasonic Imaging #Radiation Dose and Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1906.05528
openalex publication_date 2019/06/13 · openalex created_date 2022/09/11 · openalex updated_date 2026/07/28
Tomographic image reconstruction is relevant for many medical imaging\nmodalities including X-ray, ultrasound (US) computed tomography (CT) and\nphotoacoustics, for which the access to full angular range tomographic\nprojections might be not available in clinical practice due to physical or time\nconstraints. Reconstruction from incomplete data in low signal-to-noise ratio\nregime is a challenging and ill-posed inverse problem that usually leads to\nunsatisfactory image quality. While informative image priors may be learned\nusing generic deep neural network architectures, the artefacts caused by an\nill-conditioned design matrix often have global spatial support and cannot be\nefficiently filtered out by means of convolutions. In this paper we propose to\nlearn an inverse mapping in an end-to-end fashion via unrolling optimization\niterations of a prototypical reconstruction algorithm. We herein introduce a\nnetwork architecture that performs filtering jointly in both sinogram and\nspatial domains. To efficiently train such deep network we propose a novel\nregularization approach based on deep exponential weighting. Experiments on US\nand X-ray CT data show that our proposed method is qualitatively and\nquantitatively superior to conventional non-linear reconstruction methods as\nwell as state-of-the-art deep networks for image reconstruction. Fast inference\ntime of the proposed algorithm allows for sophisticated reconstructions in\nreal-time critical settings, demonstrated with US SoS imaging of an ex vivo\nbovine phantom.\n