In recent years, the ‘Vision, Learning and Optimization’ research group led by Thomas Pock at Graz University of Technology has developed ‘variational networks’ (VNs), which combine the advan-tages of traditional mathematical variational methods with deep learning methods. In collaboration with Florian Knoll at the NYU School of Medicine, these VNs were applied to the problem of MRI reconstruction. Kerstin Hammernik’s dissertation and the article written in this context, which has been cited nearly 600 times since 2018, serve as the basis for international research projects such as cooperation with Facebook AI Research on the fastMRI project. As an open-source project, fastMRI has published its data, models and code so that other researchers can build on their work and contribute new ideas. This open approach should speed progress toward clinical implementation and lead to new ways to use AI to accelerate MRI scans.
Deep Learning für MRT Rekonstruktion, from 2018
Video: FacebookAI
Project partners: Institute for Computer Graphics and Vision/Graz University of Technology, NYU Grossman School of Medicine, Facebook AI Research Funded by National Institutes of Health (NIH) scholarships/grants: R01EB024532, P41EB017183 and R21EB027241
With thanks to Kerstin Hammernik and Florian Knoll for facilitating the program on site.
