MR Image Reconstruction From Highly Undersampled k-Space Data by Dictionary Learning

By Saiprasad Ravishankar, Yoram Bresler. IEEE Transactions on Medical Imaging, 2011

1,069 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computational Mechanics, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/729fac2c883e1cf3c427c18f3d557b3d3383a92f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Computational Mechanics

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

The Little Engine That Could: Regularization by Denoising (RED) (2016, neutral): “This has been the logical path that has led to contributions such as [24, 25, 26, 27, 28, 29] and many others.”
Learning Sparsifying Transforms (2013, neutral): “(6) This problem minimizes the transform domain residual.”
DAGAN: Deep De-Aliasing Generative Adversarial Networks for Fast Compressed Sensing MRI Reconstruction (2018, methodology): “We compared our DAGAN model with conventional CS-MRI methods including TV [16], SIDWT [65] and RecPF [18], and the state-of-the-art approaches including DLMRI [26], PBDW [23], PANO [66], Noiselet [67], BM3D [50] and DeepADMM [36].”
Artifact Suppressed Dictionary Learning for Low-Dose CT Image Processing (2014, positive): “They concerned undersampled MRI image reconstruction [32], resolution enhancement [33], interior tomography [34], DL constrained iterative LDCT reconstruction [35], 3-D medical image denoising [36], few-views tomography [9]–[11], [15]–[17], [37], spectral CT [38], and abdomen LDCT image processing [”
A Perspective on Deep Imaging (2016, background): “More recently, dictionary learning, which is a contemporary machine learning approach, was adapted for MRI and CT image reconstruction [27], [30], [31].”
Deep Magnetic Resonance Image Reconstruction: Inverse Problems Meet Neural Networks (2020, neutral): “Before deep learning became popular in MRI reconstruction, there had already been quite a few learning-based reconstruction methods, where the prior models for CS reconstruction are learned from some training data (e.g., [34–37]).”
Robust Compressed Sensing MRI with Deep Generative Priors (2021, background): “Classical approaches impose sparsity in a well-chosen basis, such as the wavelet domain [58], or apply shallow learning that leverages low-level redundancy in the images [68, 69, 89].”
Deep ADMM-Net for Compressive Sensing MRI (2016, methodology): “Dictionary learning methods rely on a dictionary of local patches to improve the reconstruction accuracy [7, 8].”
A Deep Cascade of Convolutional Neural Networks for MR Image Reconstruction (2017, methodology): “We show that under the Cartesian undersampling scheme, our CNN approach is capable of producing high-quality reconstructions of 2D cardiac MR images, outperforming DL-based MRI reconstruction ( DLMRI ) [17].”
Self‐supervised learning of physics‐guided reconstruction neural networks without fully sampled reference data (2019, background): “ objective function for the reconstruction. Possible choices for the regularizer include total variation (10,45,46), 𝓁 1 -norm of wavelet coefficients (4,8,47), sparsity in adaptive transform domains (9,48), and more recently neural networks (27,28,33). The image recovery is then formulated as an o”
Interactive reconstruction of Monte Carlo image sequences using a recurrent denoising autoencoder (2017, methodology): “where each gradient ∇(·) is computed using a High Frequency Error Norm (HFEN), an image comparison metric from medical imaging [Ravishankar and Bresler 2011].”
Compressed Sensing MRI Reconstruction Using a Generative Adversarial Network With a Cyclic Loss (2017, methodology): “Earlier work using DL in CS-MRI is using image patches to train dictionary [17]–[19], which may suffer from redundant atoms and longer running time.”
Low-rank and Sparse Matrix Decomposition for Accelerated Dynamic MRI with Separation of Background and Dynamic Components (2014, methodology): “MRI presents favorable conditions for the application of CS, since (a) medical images are naturally compressible by using appropriate sparsifying transforms, such as wavelets, finite differences, learned dictionaries (4) and many others, and (b) MRI data are acquired in the spatial frequency domain ”
ADMM-CSNet: A Deep Learning Approach for Image Compressive Sensing (2020, methodology): “sparsity [16], [17], [18], [19], dictionary-based sparsity [20],”
KIKI‐net: cross‐domain convolutional neural networks for reconstructing undersampled magnetic resonance images (2018, methodology): “Dictionary learning MRI (DL-MRI) updates adaptive dictionaries by alternating back and forth between an image domain and the k-space.(14) This image-adaptive sparsifying algorithm can represent better sparsification of images compared with CS-MRI, owing to dictionaries that are learned from the imag”
Learning a Variational Network for Reconstruction of Accelerated MRI Data (2017, background): “Early attempts to use machine learning for MRI reconstruction were based on dictionary learning (37,38) 42) learned destreaking on CT images and then ﬁne-tuned the learning on MR data to remove streaking from radially undersampled k-space data.”
Magnetic resonance image reconstruction from undersampled measurements using a patch-based nonlocal operator (2014, background): “Dictionary trained from the intermediate reconstruction or fully sampled reference images have also been introduced for MRI reconstruction (Chen et al., 2010; Qu et al., 2012; Ravishankar and Bresler, 2011; Ning et al., 2013).”