Masking out background for SSIM evaluation

[briannghiem]

Hello,
Could you please share the code for the post-hoc background-masking algorithm that was described in the 2020 fastMRI paper (Section III.B - Quantitative Results)? I'm looking into ways of handling background noise when evaluating image quality metrics and wanted to know how your group approached this.

Best,

[mmuckley]

Hello @briannghiem, this workstream was led by Bruno Riemenschneider, who was a postdoc at the time and has long since left. As stated in the supplemental material, first the mask was calculated, and then all images in the challenge set were manually inspected to verify that the masking did not remove large anatomical regions.

As far as I can tell, the code was not shared on our (Meta) side and so I am unable to reproduce the masking algorithm.

[briannghiem]

Thank you for letting me know!