NYU Langone and FAIR fastMRI Dataset
fastMRI (NYU Langone Health and Meta FAIR)
Raw MRI scanner data for over 1,500 knees and 6,970 brains, built for faster scans.
Deidentified raw k-space measurements and DICOM images of over 1,500 knees and 6,970 brains. Raw k-space lets researchers test algorithms that reconstruct images from fewer measurements, with the aim of shortening MRI exams. The dataset is expected to grow annually.
Access
- Cost
- FreeNo charge to download; hosted in an AWS S3 bucket.
- License
- Open licenseMIT License
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadHDF5, DICOM
What's in it
- One row is
- One MRI scan, with raw k-space and reconstructed images
- Covers
- Not geographic
- Years
- Not stated
- Updated
- Annual
- Size
- Over 1,500 knee scans and 6,970 brain scans
Quality
Rare large public raw MRI data with a well-known reconstruction challenge. Raw data demands imaging expertise.
- Single-institution source
- Large files
Connects to
- UCSF Prostate MRI Dataset: Different MRI anatomy with clinical reads
- The Human Connectome Project: Brain imaging at research scale
Questions it could answer
- How much undersampling can reconstruction tolerate before diagnostics suffer?
- Do models trained on knees transfer to brains?