Multi-band Multi-echo EPI artifacts

, ,
1

Dear Community,

I recently encountered some artifacts in our data and could not attribute them to any specific source.

I’m collecting multi-band multi-echo data (described in this thread) with the following parameters: MB 4, GRAPPA 2, TR-1090 ms, 3.2 mm voxels, 7/8 partial Fourier, TE=13, 27.6, 42.2, 56.8, 71.4 ms. I have both phase and magnitude data.

For some of the subjects, but not all, I noticed very clear artifacts in the form of “warped checkerboard”. When they appear, they appear across several slices, raising the concern for a multi-band artifact. Here you can see the same 4 slices at different times points and echoes, each time with an artifact in a different slice:

image
image1176×287 52.6 KB

Sometimes, a distinct spatial frequency is apparent:
image

The specific case you see here is especially confusing to me, as these relatively rare artifacts become extremely prevalent towards the end of the resting-state scan, as you can see in the fmriprep report (version 25.1.4):

image
image1043×752 135 KB

On the left is an example coronal slice from early in the scan, and on the right from late in the scan:
image

These artifacts are not the result of any processing - they appear in the raw dicom files. Since we’re still collecting data from new participants, I’m worried that this might be some hardware issue we’re missing.

Any ideas on where to look for the root of the issue would be highly appreciated!

Roey

P.S.
I have not yet looked into potential tools like MARSS.

2

Hi Roey,

As you note, these look like acquisition rather than processing artifacts. There are several possibilities. Given the temporal nature of the artifacts and straight lines across images, my first guess would be that there is undesired RF noise in the scanner room. For example, maybe you have some peripheral equipment (or light bulbs) that are periodically leaking electronic signal. Another source of RF noise might be the seal on the scanner room door being broken during a scan (i.e. someone entering the scan room mid scan). If it might be something systemic with the scanner, maybe check other fMRI scans collected on the same days as the artifacts you are seeing.

Hope this helps

Dan

2 Likes
3

Hi Dan,
Thanks so much for your detailed and quick response.
We’re now looking into possible sources of RF noise. One reason I thought this could be a scanner issue is that I’ve noticed several cases in which the artifacts were much more prevalent towards the end of the scan.
I’ll post an update when we’ve identified the issue.
Many thanks again,
Roey

4

Hi Roey,

For the specific case of the “distinct spatial frequency”: this is a clear case of a “spike” in the signal in k-space, which results in high intensity at a specific frequency and direction. This is normally produced by “arcing” in the hardware.

When I was at NYU, there was some months in which we had these types of artifacts, although they didn’t happen for minutes at a time. In that case, we nailed it down to the air being too dry (although it was within the scanner manufacturer’s specs). The solution was to add a commercial-grade humidification system in the supply line going into the equipment room and scanner room.

Given that you have long periods of time of those artifacts, the first thing I would check is to see if during all those epochs the artifacts all have a fixed spatial frequency, or it’s something else.

The other artifacts could be some errors in the reconstruction. It could be that, as the scan goes on, the initial coil calibration is no longer valid, either because of motion or because of hardware heating up, and you get those artifacts you got in the first four images. After all, you are running with an R of 8 (4x2), which is a little aggressive, IMHO. Or, as Dan suggested, maybe there is some other source of RF noise in the room.

Things to try:

  • The first thing I would try is to collect data on a phantom with the same protocol (including number of volumes and number of runs). That would rule out motion being the cause.
  • Try running with a less-aggressive acceleration (either lower MB factor, or removing the in-plane acceleration). You’ll have to change the TR (and TE, if you change the GRAPPA). If the source of artifacts is RF noise from the room, it really doesn’t matter the acceleration: it should be picked up even without any acceleration or Multi-Echo.
  • Try using a different head coil (you might have to change the acceleration). If data collected with a different coil doesn’t have this artifact, then you have a problem with the original coil (maybe it becomes too loose as the scan progresses?).

Good luck with your hunt. Solving this intermittent issues is really hard…

-Pablo

2 Likes
5

Dear Pablo,

Thank you so much for your sympathy, as well as for the helpful tips.
Indeed, we found this artifact when running on a phantom as well, and also when removing GRAPPA and when changing the multi-band factor to 2. It is also apparent using a different head coil.

We did not consider the humidity yet - thanks for this suggestion.

Many thanks,
Roey