As an MRS user who has tried both OSPREY and GANNET, I have found that they share some similarities in data analysis. However, I also feel a little confused when choosing between them. Could you please provide some advice on the scenarios for using each software?
I have data scanned with MEGA-PRESS sequence for GABA+ concentration and also data with PRESS sequence for Glx.
Should I use GANNET for GABA+ analysis and OSPREY for PRESS analysis? I compared the results of the two softwares on the same MEGA-PRESS sequence targetting for GABA+. However, I didn’t get identical results, which I suspect is due to slight differences in the fitting models. Am I correct in my understanding?
Your advice would be greatly appreciated! Thank you very much!
You are correct in that GAnnet and Osprey will give you different results with your MEGA-PRESS data. You can read our paper to learn more about how all software toolkits differ in their outcomes.
While you may use Gannet for your MEGA-PRESS data and Osprey for your PRESS data, my recommendation would be to use Osprey for both. One of Osprey’s advantages over Gannet is that it can fit more types of MRS datasets. For the sake of keeping things as similar as possible, I would use Osprey for both if both types of data are part of the same study and are to be reported together.
I agree with your suggestion that Osprey would be the appropriate choice for ensuring consistency in the analysis.
However, I decided to initially run my dataset through Gannet due to the fact that some previous researchers, who conducted similar studies also used Gannet for their analysis. Unfortunately, I have encountered an issue that is bothering me: The patterns that emerge after applying Gannet (such as the mean GABA_conc_Cr in group A > group B) differ from the patterns observed after using Osprey (where the mean GABAPlus scaled in Cr in group A < group B). Each group consists of only around 10 individuals now, but these inconsistencies are causing me some concern.
For both methods, I ensured that the spectral alignment was set to RobSpecReg, and in Osprey, the opts.fit.coMM3 option was set to 1to1GABA. I would greatly appreciate your advice on how to interpret these contrary results.
Additionally, I noticed that when assessing processing quality, the fluctuation in Cr frequency observed in Osprey appeared to be minimal, whereas in Gannet, the variation in water frequency seemed to be unacceptable. I uploaded two screenshot to show it.
Does this imply that Gannet employs a more stringent artifact removal procedure?
Going forward, could you please move and recopy your query/this thread out of this topic and into a new and separate topic in Gannet - MRSHub Forum? This will help others find your query more easily if they have a similar concern.
To answer your question:
I would need to see your concentration results—i.e., the GannetFit and OspreyFit results—to better determine the extent of the difference. The GannetLoad and OspreyLoad results you’ve shared tell me that both toolkits have processed the data well.
A recent report from the Osprey team showed that the 3to2MM option is better for reproducibility for the GABA-edited signal.
Also, 10 subjects are quite low in determining reliable differences, but that is based on what your study is investigating, the data quality, and how big of an effect you expect. So, the results differences between Osprey and Gannet might not be so surprising.
Thanks for your kind. I would copy this topic to the Gannet- MRSHub Forum and hope it could be helpful for other MRS researchers too. I also tried 3to2MM option in re-analysis and to do the comparison.
