fastMRI: An Open Dataset and Benchmarks for Accelerated MRI

By Jure Zbontar, Florian Knoll, Anuroop Sriram, Matthew Muckley, M. Bruno, Aaron Defazio, Marc Parente, Krzysztof J. Geras, Joe Katsnelson, Hersh Chandarana and 13 more. arXiv.org, 2018

1,072 citations recorded by the source corpus.

Accelerating Magnetic Resonance Imaging (MRI) by taking fewer measurements has the potential to reduce medical costs, minimize stress to patients and make MRI possible in applications where it is currently prohibitively slow or expensive. We introduce the fastMRI dataset, a large-scale collection of both raw MR measurements and clinical MR images, that can be used for training and evaluation of machine-learning approaches to MR image reconstruction. By introducing standardized evaluation criteria and a freely-accessible dataset, our goal is to help the community make rapid advances in the state of the art for MR image reconstruction. We also provide a self-contained introduction to MRI for machine learning researchers with no medical imaging background.

Subfields: Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/65d3e7fa7b452611aa729ac789d4373d6ee505b0

How researchers see this paper

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

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

EvalAI: Towards Better Evaluation Systems for AI Agents (2019, background): “7.4 fastMRI. fastMRI [35] is a collaborative research project between Facebook AI Research (FAIR) and NYU Langone Health to investigate the use of AI to make MRI scans upto 10 times faster.”
Deep-Learning Methods for Parallel Magnetic Resonance Imaging Reconstruction: A Survey of the Current Approaches, Trends, and Issues (2019, background): “While the momentum in initiatives for public releases of raw k-space data is growing [79], the number of available data sets is still on the order of hundreds and limited to very speciﬁc types of exams.”
End-to-End Variational Networks for Accelerated MRI Reconstruction (2020, methodology): “These random masks are identical to those used in [18].”
Prior Image-Constrained Reconstruction using Style-Based Generative Models (2021, neutral): “For the inverse-crime study, StyleGAN2 was trained on a composite brain MR image dataset consisting of a total of 200676 T1 and T2 weighted images of size 256×256 from the fastMRI initiative database (Zbontar et al., 2018) and 866 images from the brain tumor progression dataset (Schmainda & Prah, 20”
Results of the 2020 fastMRI Challenge for Machine Learning MR Image Reconstruction (2021, methodology): “The first consists of the release of a large data set of raw k-space and DICOM images [25], [26].”
Diffusion models in medical imaging: A comprehensive survey (2022, methodology): “The overall results over the single coiled fastMRI [94] knee dataset and private liver MRI dataset indicate the superiority of this approach over conventional SOTA un/self-supervised learning schemes in terms of SNR and Contrast-to-Noise Ratio (CNR) metrics.”
Transformers in Medical Imaging: A Survey (2022, methodology): “MRI 2D fastMRI [304] IXI [303] PSNR, SSIM, NMSE”
Solving Inverse Problems in Medical Imaging with Score-Based Generative Models (2021, methodology): “We simulate measurements by computing the k-space with Fast Fourier Transform, and follow [22, 23] to undersample the k-space with an equispaced Cartesian mask.”
Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models for Inverse Problems through Stochastic Contraction (2021, methodology): “Results of the MR reconstruction task: (a) TV [3], (b) UNet [41], (c) score-POCS [6] using 1000 steps starting from T = 1, (d) proposed method (CCDF) using 20 steps from t0 = 0.”