Using a RPE fieldmap image from a Philips scanner: specifying directions

[emilymharriott]

@Steven

You can see in Table 1 the importance of each QC metric in predicting whether an image would pass manual quality review (i.e., the gold standard). Neighbor correlation was at the top of the list and is likely an index of lots of underlying quality items (such as motion).

Got it!! It seems like I’m particularly looking for the neighboring DWI correlation (want it to be high), maximum relative translation (want it to be low), and number of outlier slices (want it to be low). Thank you for pointing me to this paper!

[Steven]

@emilymharriott

Should I still take the combo approach going forward, because using real data is better?

This isn’t necessarily always true. In fact, the Syn paper (https://www.frontiersin.org/articles/10.3389/fninf.2017.00017/full) found that the the fieldmapless method could work as well or as better as fmaps that were not acquired optimally (e.g., motion). I think synb0 will work better than the Syn method, when it is finished being integrated into qsiprep.

That being said, I thought the real fmap looked the best, but I haven’t done any overlays with T1 images as you and your postdoc might have.

[emilymharriott]

@Steven

I think synb0 will work better than the Syn method, when it is finished being integrated into qsiprep.

Super helpful, thank you!

[emilymharriott]

Thank you all so so so unbelievably much @Steven @mattcieslak @jsein !!! I really appreciate your insight & speedy replies!!!