fastMRI

Matching contrasts in Knee Images

Hi. In the FastMRI paper, it is written that there are 798 scans with PDFS and 796 scans with PD. Are the 796 scans coming from the same cross-sections of the same patients? Is it possible to have the PD and PDFS version of the same cross sections? Or is it the case that only one contrast was chosen for one patient?
As a second question, are the training, validation and test knee images fully sampled?

Hello @Ridvan_Yesiloglu, we cannot match the cases. For each patient NYU only acquired one fully-sampled scan due to time constraints. This was either the PD or the PDFS scan.

You can download fully-sampled training and validation images from the NYU web site. The test data is of course only subsampled so that we can maintain the integrity of our public leaderboard.

Thank you @mmuckley. I am in need of a multicoil, multicontrast data, which I can match the cases but I am having a hard time to find. Is there such a dataset in MRI?
Sorry, it might be out of this forum’s topic but I am in need of an advice.

Hello @Ridvan_Yesiloglu, unfortunately I’m not aware of any large data sets for this on the scale of fastMRI.