fastMRI: A Publicly Available Raw k-Space and DICOM Dataset of Knee Images for Accelerated MR Image Reconstruction Using Machine Learning

Abstract

A publicly available dataset containing k-space data as well as Digital Imaging and Communications in Medicine image data of knee images for accelerated MR image reconstruction using machine learning is presented.

Figure 1:

Coronal proton density−weighted images with fat suppression (left) and without fat suppression (right). Both images were reconstructed from fully sampled k-space data using a sum-of-squares combination of component coil images.

Figure 2:

Examples of binary sampling masks (white = included, black = omitted) for pseudorandomly undersampled k-space data with fourfold acceleration (left) and eightfold acceleration (right). The overall acceleration factor is set randomly either to four or to eight (representing a fourfold or an eightfold acceleration, respectively), with equal probability for each example. The undersampling mask is then generated by first including some number of adjacent low-frequency k-space lines to provide a fully sampled central region of k-space. When the acceleration factor equals four, the fully sampled central region includes 8% of all k-space lines; when the acceleration factor equals eight, 4% of all k-space lines are included. The remaining k-space lines are included at random, by drawing samples from a uniform random distribution with the probability set such that the correct number of total k-space lines is achieved.