We are making available training and test data set used in our related publications on machine learning for reconstruction of accelerated MRI data. For details, see the related publications below.
The 100 cases include 20 cases from each of the following five sequences:
- coronal spin density weighted with fat suppression
- coronal spin density weighted without fat suppression
- axial T2 weighted with fat suppression
- sagittal T2 weighted with fat suppression
- sagittal spin density weighted
The data are organized slice by slice, and consist of two files:
The rawdata and some additional metadata variables (the acquisition includes phase oversampling and a rectangular field of view, see example matlab reconstruction script)
ESPIRiT coil sensitivity maps and a reference reconstruction
In addition, in separate folders associated with the data from each sequence, we include the parallel-imaging subsampling used to conduct some of the experiments in the related publications.
Learning a variational network for reconstruction of accelerated MRI data.
Magn Reson Med. 2017 Aug;78(2):565-576. doi: 10.1002/mrm.26977
Assessment of the generalization of learned image reconstruction and the potential for transfer learning.
Magn Reson Med. 2019 Jan;81(1):116-128. doi: 10.1002/mrm.27355
Please cite these works if you are using the 100 knee MRI cases data set in your research.
ESPIRiT–an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA.
Magn Reson Med. 2014 Mar;71(3):990-1001. doi: 10.1002/mrm.24751
Questions about this resource may be directed to Florian Knoll, PhD.
Learn More about This Research
This is a joint project with Thomas Pock’s vision, learning, and optimization group at the Institute of Computer Graphics and Vision at Graz University of Technology.
- Interview with the reserach team, plus video slides in “MRM Highlights” (June 2018)
- Video of a presentation delivered at the i2i Workshop in New York (October 2016)