In LCModel analysis, it is preferable that the TE of the data and the TE of the Basis set almost match, but wouldn’t it be a problem if the difference was only 1 ms? I would appreciate it if you could tell me the supporting documents.
If your basis sets almost match, your results are likely to be only almost as good as they could be . It’ll work, and the differences are likely to be very subtle – but may nonetheless affect the reliability of separation, particularly for highly overlapping peaks. If you’re using a generic basis set, consider that other factors (pulse timing and shapes) may differ too.
There are several tools available for creating basis sets to more closely match your sequence, so that would be the best way forward. MRSCloud may be the easiest option for common sequences, FSL MRS, VESPA and FID-A all provide this functionality too.
Otherwise, pay close attention to your fit residuals and baseline model.
Thank you for your reply.
Subtle differences in TE are allowed. Is there any literature regarding this?
I don’t think I said that
The experts’ consensus recommendation states:
the simulation parameters should match the in vivo acquisition (pulse sequence, field strength, echo time and optionally the RF pulse shapes and durations).
In any case, the basis set is a model – it’s not expected to be perfect, but naturally we’d prefer the best model available. If you choose a slightly less suitable model you will still get results, but they may be slightly less reliable. In many common cases, 1ms will not make a huge difference – but in some cases with difficult-to-separate metabolites it might. Whether or not this compromise is acceptable very much depends on context.
If unsure, the consensus recommendations are a good place to start.