Data augmentation for deep learning based
             accelerated MRI reconstruction with limited data

                 Zalan Fabian 1 Reinhard Heckel 2 Mahdi Soltanolkotabi 1

              Abstract                tients. Due to the subsampling, there are fewer equations
    Deep neural networks have emerged as very suc-       than unknowns, and therefore the signal is not uniquely iden-
    cessful tools for image restoration and reconstruc-     tifiable from the measurements. To overcome th ...