Image Reconstructors

Image Reconstructors#

Sub-package containing image reconstruction wrapper routines.

Building blocks#

DeepMR contains a convenient builder routines to automatically initialize the MR encoding operator for different reconstruction problems (e.g., Cartesian vs Non-Cartesian, single- vs multi-channel, single- vs multi-contrast).

This can be used inside conventional reconstruction algorithms, or inside neural network architectures.

deepmr.recon.EncodingOp

Prepare MR encoding operator for Cartesian / Non-Cartesian imaging.

Classical Image Reconstruction#

DeepMR contains convenient wrappers around generic standard reconstruction.

The provided routine can perform zero-filled or iterative reconstruction with Tikonhov, L1Wavelet or Total Variation regularization for both Cartesian and Non-Cartesian (single-/multi- contrast/channel) data, depending on input arguments.

deepmr.recon.recon_lstsq

Classical MR reconstruction.