Suggestion about ghosting in BOLD images

1

Dear all,

I’m a newbie in fMRI and I collect some pilot fMRI data recently. I found there are apparent ghosting in bold images like this

image
image1749×1331 266 KB

I’m not sure how bad it is. Will this ghosting influence my further fMRI analysis severely? How should I deal with it? Thank you!

2

some possibly important scan parameters

image
image1081×1208 99.5 KB

image
image1244×767 63.8 KB

3

Hi,
Ghosting is inherent to echo planar imaging (EPI) acquisition. Visually, just from this image it is difficult to tell if it is worse than usual, but it seems normal to me.
It is possible to assess this artefact quantitatively with metrics such as Ghost to Signal Ratio implemented in tools like MRIQC:

https://mriqc.readthedocs.io/en/latest/iqms/bold.html#mriqc.qc.functional.gsr

Looking at your parameters, I have some comments:

  • how many slices did you use? It looks like you have a TR of 1s, with slices of 2mm. Do you use any kind of acceleration? (in-plane (GRAPPA) or multi-slice?)
  • You use a flip angle of 90 degrees which is not optimal for such a short TR, the Ernst Angle would be about 62 degrees. (taking a T1 of GM of 1300ms). Some studies recommend to lower even more the flip angle to reduce the physiological noise (Gonzalez-Castillo et al. 2011)
4

Hi jsein,

Thanks you! I have run the MRIQC check and here is the group level (2 participants) bold report. Not sure whether the gsr here is acceptable or not.

image
image3709×1674 442 KB

As for the questions:

  • We used 65 slices and yes, we used multi-slice acceleration.
  • Thanks for your suggestion. Should I consider other parameters if fine-tuning to smaller flip angle? Sorry that I know very little about the physical principles of MRI.

Thanks for your help!

5

As @jsein says, some ghosting is inevitable with EPI, and it gets more prominent (and more frequent) with accelerated sequences. practiCal fMRI: the nuts & bolts: Use of split slice GRAPPA (aka Leak Block) for SMS-EPI reconstruction has background information and examples; look especially at the “COASTER” rings.

Judging how much ghosting is “too much” is not straightforward, and (as usual) depends on multiple considerations, including the type of analysis you’re planning. Definitely ask your colleagues, imaging center staff, physicists, etc. for their opinion of your pilot images (and acquisition parameters). Looking at other datasets can also help to train your eye, e.g., OSF.

If at all possible, include some control analyses (tasks known to produce strong, consistent responses in the brain areas of most interest to you) in your piloting, so you can confirm that the acquisition produces reasonably clear signal.

6

Thanks for your explanation and suggestion! I will check it.