Mohammad Zalbagi Darestani1 and Reinhard Heckel1,2
1Electrical and Computer Engineering, Rice University, Houston, TX, United States, 2Electrical and Computer Engineering, Technical University of Munich, Munich, Germany
1Electrical and Computer Engineering, Rice University, Houston, TX, United States, 2Electrical and Computer Engineering, Technical University of Munich, Munich, Germany
Our work shows that a close performance to trained neural networks can be achieved without training using an un-trained network for accelerated MRI. We further show that untrained networks have another advantage, that is, naturally generalizing better to un-seen samples.
ConvDecoder architecture. It is comprised of up-sampling, convolutional, ReLU, batch normalization, and linear combination layers.
Sample reconstructions for ConvDecoder, TV, U-net, and the end-to-end variational network (VarNet) for a validation image from multi-coil knee measurements (4x accelerated). The second row represents the zoomed-in version of the first row. ConvDecoder and the end-to-end variational network (VarNet) find the best reconstructions for this image (slightly better than U-net and significantly better than TV). The scores given below are averaged over 200 different mid-slice images from the FastMRI validation set.