Score-based diffusion models for accelerated MRI

By Hyungjin Chung, Jong Chul Ye. Medical Image Anal., 2021

590 citations recorded by the source corpus.

Score-based diffusion models provide a powerful way to model images using the gradient of the data distribution. Leveraging the learned score function as a prior, here we introduce a way to sample data from a conditional distribution given the measurements, such that the model can be readily used for solving inverse problems in imaging, especially for accelerated MRI. In short, we train a continuous time-dependent score function with denoising score matching. Then, at the inference stage, we iterate between the numerical SDE solver and data consistency step to achieve reconstruction. Our model requires magnitude images only for training, and yet is able to reconstruct complex-valued data, and even extends to parallel imaging. The proposed method is agnostic to sub-sampling patterns and has excellent generalization capability so that it can be used with any sampling schemes for any body parts that are not used for training data. Also, due to its generative nature, our approach can quantify uncertainty, which is not possible with standard regression settings. On top of all the advantages, our method also has very strong performance, even beating the models trained with full supervision. With extensive experiments, we verify the superiority of our method in terms of quality and practicality.

Subfields: Mathematical Physics, Radiology, Nuclear Medicine and Imaging, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/ff85a9d7182063bb71b47ab239d662fd2975c4fc

How researchers see this paper

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

In Mathematical Physics

In Radiology, Nuclear Medicine and Imaging

In Statistical and Nonlinear Physics

Citations in the corpus

Diffusion Posterior Sampling for General Noisy Inverse Problems (2022, methodology): “Due to its intractability, one often resorts to projections onto the measurement subspace (Song et al., 2021b; Chung et al., 2022b; Chung & Ye, 2022; Choi et al., 2021).”
Diffusion Models: A Comprehensive Survey of Methods and Applications (2022, background): “Computer Vision Image Super Resolution and Inpainting [119],[174],[170],[129],[172], [79],[12],[154],[34],[191],[32] Semantic Segmentation [11],[5] Anomaly Detection [217],[62] Point Cloud Completion and Generation [248],[133],[138],[126] Video Generation [76],[81],[230],[238] Natural Language Proce”
Diffusion models in medical imaging: A comprehensive survey (2022, methodology): “Generative models, specifically diffusion models, can be used as an alternative to these model-based priors since they can provide a more accurate representation of the data distribution [47, 48].”
A Survey on Generative Diffusion Models (2022, methodology): “They have been successfully used in super-resolution [58], [160], classiﬁcation [162], and noise robustness [161], [163].”
Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models for Inverse Problems through Stochastic Contraction (2021, methodology): “(a) Corrupted measurement, (b) ILVR [5], score-SDE [34], and score-MRI [6], respectively, for each task.”