Reconstruction error on VE11C

[josephholtrop]

Hello,

A few months ago, we started getting an image reconstruction error on one of our Prisma's (VE11C) when we ran any of the CMRR multiband sequences. After shimming the scan fails to begin with the system giving us an image reconstruction error. We tried retrying the scan, rebooting, reinstalling the sequence, installing the newest available version (R2017pre5), and replacing the MARS. This error only occurs when the multiband factor is set to 2 or higher. Additionally, we have no issues running the Siemens multiband EPI. We ran out of ideas to fix this and were able to move our CMRR scans to other scanners to minimize disruptions.

Yesterday, we started getting the same reconstruction message on another scanner (Skyra VE11C).

Any advice on what could be causing our issues?

[rwmair]

Hi Joseph, What site/institution are you at, and how many VE11C scanners do you have CMRR R017pre? installed on?

When you say the scan fails to begin after shimming - it makes me wonder if this actually an image preparation error, not a reconstruction error. As nothing has actually run to that point.

By any chance are you running with the 64ch coil? I've seen a few instances in recent times at different sites where shimming with the 64ch coil fails. Its apparently a known bug with the 64ch coil and VE11C. The shimming fails and the scan then refuses to start, with errors that look like image-prep or recon failures. Tho our site and the other where I've seen this have had VE11C and the 64ch coil for years, this problem seems to be occurring now much more frequently than before - for an unknown reason. The solution (from Siemens) is to switch to "adjust with Body Coil" on the System/Adjustments page, shim again and then the scan will run.

I don't know if any of this will address your problem, but figured I'd throw it out there in case its an issue with the shim routine failing.

[josephholtrop]

I am at St. Jude Children's Research Hospital. We were running R016a on 3 3T scanners (2 Prisma, 1 Skyra). R015pre was installed on one Prisma when we started getting errors to see if that would fix anything (it did not). Currently we are down to 1 Prisma that is able to run CMRR multiband.

I agree that this seems like a image preparation error. In the log we see messages about not being able to load 'libCMRR_MBIcePrg2D.so'. We have verified that this file exists in the ice folder. I should also note that with the 'emptyice' switch turned on the sequence runs and we are able to acquire a dataset ( and reconstruct it on our 1 working scanner), but retro recon fails in the same way.

We are using the 20 channel head/neck coil on all our scanners. We will try running with body shim.

[eauerbach]

Usually this arises due to a problem with the MaRS, specifically the GPU(s) in the MaRS. Replacing the MaRS once unfortunately wouldn't rule it out, since it is likely that most of the replacement VE11C-era MaRS still in inventory have been "refurbished", maybe multiple times, probably not perfectly. Sometimes it takes a few replacements to get a fully functioning one.

Anyway, a quick test for this is to delete the file C:\Medcom\MCIR\Med\config\eja_hardware.ini. This will cause the MB reconstruction to ignore any GPU. (Note that it is a temporary test, the file is regenerated automatically when the MaRS reboots.)

Without GPU the recon will be very slow, but if it magically starts working again you know you have a GPU issue. On VE11C the product sequences do not use the GPU heavily, and IIRC they will never use two GPUs if you have them. So a failure of a second GPU will only cause errors with the MB recon.

[rwmair]

Hi Joe, between what you and Eddie wrote, I think its safe to ignore my suggestion related to the shimming method. 20ch coils should not have this shim issue. Following Eddie's suggestions to check if you have non-functional GPU's in the MARS seems the way to go.

[josephholtrop]

After deleting C:\Medcom\MCIR\Med\config\eja_hardware.ini, we are still seeing the same errors on both our Skyra and one of our Prismas. We are working on getting another new MARS in to try that out. Any other advice that might help us narrow down what our issue is?

[rwmair]

Is there any chance you can swap the MARS units temporarily between the two Prisma scanners? If the error goes with the MARS unit, then that should identify the culprit, I would think. If the Skyra has the same model MARS, you could try the working-Prisma MARS in there also just to confirm.

[josephholtrop]

@rwmair We have discussed doing that swap. We are hesitant to do anything that would impact our one working system at this point, as we would that would severally impact our ongoing studies if that currently working system did not come back online correctly. We may decide to do that swap if no other good options come up.

[PauleMarcouxValiquette]

Hi, I was running the CMRR Multiband EPI sequence on VE11C and about 25 images in, the sequence stopped, and I got the following error message (Image reconstruction failed. MeasUid 36)

We have never had this issue before, and this was not a long sequence – about 8 minutes. After a few tests, I determined that it was the multi echoes (Contrast =3) that were causing the issue since I had no problem running the sequence for MB 1, 2 and 3 with 1 contrast. I restarted Meas-Recon tool and did a full shut down and Restart of the MRI and that didn’t change anything.

Any idea how to fix this?

[eauerbach]

Again that is usually an issue with the MaRS, as increasing the number of echoes increases the processing demand.

Depending on the exact error it could also be a problem with the image database on the host, if the host cannot accept images fast enough. In that case rebuilding the image database might help.

[zbenyouss]

Are there any updates on this? We are having the same issue on our scanner. After 30 volumes we get measurement failed, and looking at the MRI error log we see "Reconstruction Failed. MeasUID 2243"