Hi - We are running into an issue with inconsistent results from the fit step of Osprey (v. 2.8.1). Results are inconsistent between users and we have been unable to determine what differs in each environment that would cause the inconsistency.
At the dropbox link are the matlab environmental version settings, the job files, the .mat files, and the output for each system’s run. The raw data is also included. Both users are launching matlab in a remote desktop environment on our cluster (through their own user profile).
We’ve confirmed that the same data is being used by each run, the same versions of Osprey, SPM12, and Widget-toolbox have been installed, and the same matlab version is being used. It looks like all the same options have been selected in the job file.
We’ve also found that results are consistent within system when comparing two recent versions of Osprey (2.8.1 vs. 2.8.5).
Are there other troubleshooting steps we can take to resolve the inconsistency? Appreciate any help solving this.
Hi @a.waters,
Thank you for reporting this issue. We are aware of some between-OS issues with the modeling and are working on a fix for that right now. I assume that this is the same issue that you are facing.
I’ll take a look at the data you shared and let you know when we have made the appropriate changes.
Best,
Helge
Thanks Helge,
In the interim, is there a specific OS that is preferred?
I am finally done with the modifications. The updated version gave me the same metabolite estimates for my test data (short TE PRESS and HERCULES) across Mac, Windows, and Unix.
The updated version can be found here (GitHub - schorschinho/osprey at Osprey2.10). Can you give that a try to see if this resolves the issues with the remote desktop environment as well?
Best,
Helge
Thanks! We ran the new version with our HERMES data. This discrepancy at the Model/Fit step has been resolved. However, we are getting an error at the Quantify step. Two different errors are returned depending on whether Quantify is applied after segmentation or if the segmentation step is skipped (see below).
Good to hear that the discrepancies are resolved. Could you post the sum, diff1, diff2 model results here?
I’ll take a look at the errors you are reporting. I see that you have ran everything through the GUI. Did you try running it through the command window as well?
Also, which MM3co model are you using?
Hi @a.waters,
I have fixed the issues. Can you give it a try and send me the model results?
Best,
Helge