Deep learning has shown tremendous success in several fields of science such as image and natural language processing. In this talk we demonstrate how these advancements can be extended to Image Reconstruction by embedding the forward operator inside a classical Convolutional Neural Network. We demonstrate some such Learned Iterative Reconstruction schemes and give state of the art results from Computed Tomography.

This presentation is part of Minisymposium “MS33 - Advances in reconstruction algorithms for computed tomography (4 parts)”
organized by: Gunay Dogan (Theiss Research, NIST) , Harbir Antil (George Mason University) , Elena Loli Piccolomini (Dept. Computer Science and Engineering, University of Bologna) , Samuli Siltanen (University of Helsinki) .

Authors:

Jonas Adler (KTH Royal Institute of Technology)

Ozan Öktem (KTH - Royal Institute of Technology)

Keywords:

computed tomography, deep learning, image reconstruction, inverse problems, machine learning