I’m wondering about the role of the option coMM3 in Osprey. I guess that it has to do with how the fitting procedure treats coediting of some macromolecule, possibly MM20, which has a peak at 3.00 ppm?
I’m currently collecting data both with the standard GABA editing sequence, with editing pulses at 1.9 ppm (ON) and 7.46 ppm (OFF), and with the MM supressed variant, with editing pulses at 1.9 ppm (ON) and 1.5 ppm (OFF).
Should coMM be set different for these two sequences and what should it be set to? Is there any other considerations for these two types of editing?
On a completely different thing, and perhaps meriting a separate topic, do you have any plans for support of functional MRS, in the sense that data can be collected during a set of conditions (“task”/“rest” blocks), such that one can separate the data for “task” and “rest” in the analysis. For example, the experiments has been set up so that subspectrum [(n-1)16+1 : n16] corresponds to “task” for even n and to “rest” for odd n. (The blocks of 16 subspectra would be 8 OFF and 8 ON, respectively, and n would typically go up to 32 or more.) It is certainly possible to sort the data after loading and modification of the MRSCont structure but it would be awesome to have it built into Osprey from start.
Thanks for reaching out. @Helge is currently examining optimal ways of modeling the 3-ppm GABA+ peak, which is clearly inappropriately modeled if only the GABA basis function is included. We know that, at the very least, there is homocarnosine in there (which we can simulate), and obviously the MM signal. There’s not a lot in the literature about how to model the MM contribution, so we’re exploring whether it’s robustly parametrizable with a Gaussian of some sort. We’re planning to publish a paper on that at some point, but for now we are not making any clear recommendations on the parametrization. In particular, we have not investigated this for MM-suppressed data. I’m not convinced at all that the MM signal is entirely removed even for perfectly stable acquisitions (i.e. zero drift), so “MM suppression” is a bit misleading in my eyes anyway - and on top of that, the actual MM signal will be enormously sensitive to the actual course of the experiment, i.e. frequency offset history. There are ways to account for that by constructing the basis sets from the frequency history (see Jan Willem van der Veen’s excellent paper), but it is less clear how the MM contribution is treated in that scenario. Certainly a very exciting study to conduct…
With regard to your second question: We have several groups that contacted us to ask whether we have plans to implement fMRS processing and modelling. It’s a very interesting and fashionable feature that, I hope, will find its way into Osprey at some point (and you’re right - it’s not “difficult” to conceive a workflow, but it requires a lot of time). At the moment I’m not sure we have the bandwidth to drive the implementation ourselves, as we don’t have ongoing fMRS research in our groups here at Hopkins. Ideally, someone with an active interest (read: funding and ongoing study) in fMRS could approach us and we can jointly find ways to move the implementation forward. Another way to get started would be for you to acquire a few test datasets, and start designing your own fMRS processing module within the Osprey framework - we would be more than happy to assist you, and of course give credit where appropriate.