Abstract
Accelerating MRI scans is one of the principal outstanding problems in the MRI research community. Towards this goal, we hosted the second fastMRI competition targeted towards reconstructing MR images with subsampled k-space data. We provided participants with data from 7,299 clinical brain scans (de-identified via a HIPAA-compliant procedure by NYU Langone Health), holding back the fully-sampled data from 894 of these scans for challenge evaluation purposes. In contrast to the 2019 challenge, we focused our radiologist evaluations on pathological assessment in brain images. We also debuted a new Transfer track that required participants to submit models evaluated on MRI scanners from outside the training set. We received 19 submissions from eight different groups. Results showed one team scoring best in both SSIM scores and qualitative radiologist evaluations. We also performed analysis on alternative metrics to mitigate the effects of background noise and collected feedback from the participants to inform future challenges. Lastly, we identify common failure modes across the submissions, highlighting areas of need for future research in the MRI reconstruction community.
Highlights
D UE to advances in algorithms, software platforms [1]–[3] and compute hardware, over the last five years there has been a surge of research of MR image reconstruction methods based on machine learning [4]–[14]
The results were better on the high signal-to-noise T2 and FLAIR contrasts compared with those on the T1 post-contrast images (T1POST)
The 2020 fastMRI reconstruction challenge featured two core modifications from its 2019 predecessor: 1) a new competition Transfer track to evaluate model generalization and 2) adjusting the radiologist evaluation to focus on pathology depiction
Summary
D UE to advances in algorithms, software platforms [1]–[3] and compute hardware, over the last five years there has been a surge of research of MR image reconstruction methods based on machine learning [4]–[14]. Research in MR image reconstruction methods has been conducted on small data sets collected by individual research groups with direct access to MR scanner hardware and research agreements with the scanner vendors. Research groups lacking large-scale data collection infrastructure face substantial barriers to reproducing results and making comparisons to existing methods in the literature. In the field of computer vision, the basic principles of convolutional neural networks (CNNs) were proposed as early as 1980 [15] and became well-established for character recognition by
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call