Dear All,

Hi,

Baseline distortion is considered an artifact…but whatabout non-flat baseline?

Which is recommended to do in TARQUIN analysis? baseline subtraction or smoothing?

Thank you

Dear All,

Hi,

Baseline distortion is considered an artifact…but whatabout non-flat baseline?

Which is recommended to do in TARQUIN analysis? baseline subtraction or smoothing?

Thank you

Hi Neda,

Tarquin uses a time-domain truncation approach to deal with (macromolecule) baseline. Essentially: based on the fact that macromolecule signal components generally decay more rapidly than metabolite signals, truncating some initial points from the time-domain FID should reduce their contribution to the processed spectrum. Section 4.3.2.1 of Poullet et al 2008 provides a good contextual overview of this (see also the cited Stanley et al 2001 and Ratiney et al 2005)

In Tarquin, the number of points to be removed from the beginning of the FID is determined by the `start_pnt`

parameter (“Start Point” in the GUI). As with many parameters, the choice of value is a tradeoff (see articles above); I believe Tarquin tries to estimate a reasonable initial value, so it’s probably safest to stick with this value unless the result you see is poor/unrealistic.

Alex.

Hi Alex,

Thank youvery much for the detailed explanation and the suggested references

Hi,

what about eddy current correction? is it a part of baseline correction?

Hi @NNeda,

No, eddy-current correction is not part of the baseline correction. Eddy-current correction is a method to determine a non-linear phase in the time-domain signal that is induced by a transient and time-dependent change in B0, which is induced by rapid switching of gradients.

“Baseline correction” happens during spectral modeling and is a somewhat misleading term, because there is not actually a correction going on. It much rather describes attempts to model slow fluctuations in the background signal, for example resulting from residual water, but also lipids are often modeled with a baseline term. LCModel does this with a spline, while Tarquin truncates the first 10% (or so ?) of the time-domain signal - see Alex’ explanation above.

I can highly recommend reading about both concepts in Robin de Graaf’s excellent MRS introduction book (https://onlinelibrary.wiley.com/doi/book/10.1002/9781119382461).

1 Like

@admin

Hi,

Thank you very much for the explanation. I just found out my mistake and I wanted to correct it that I saw your answer

Does any procedure in TD necessarily mean prior to model fitting??

I think I am a little bit confused by thses two words: k-space and spatial…What I know it that k-space means after FT (frequency domain) but spatial means we need to have an inverse FT(?) and it’s spatial domain and image space(?)…do k-space interpolation must be different than spatial interpolation(?)…zero filling is which of them?

T would really appreciate your reply

Any operation in TD means before model fitting but operations before model fitting can be in FD too(?)…

There is sentence : operations before model fitting is called preprocessing but,

operations after FT is called post processing…

By this definition an operation can be called both preprocessing and post…?

These are a lot of different concepts you’re bringing up here. Are you trying to solve a specific problem or are these questions that occur while you are reading textbooks?

I’ll try a few responses, but I think you’ll need to provide a little more context for your questions:

- The terminology is pretty difficult and people use different words for the same things. To me, pre-processing is every operation that happens before fitting, but other people call this post-processing. Jamie’s analysis consensus paper calls it pre-processing, so I think there’s good reason to stick with this definition.
- Yes, pre-processing operations can be done in both the frequency and the time domain. For what it’s worth, modeling/fitting can also be done on the time- or the frequency domain signal.
- I’m not sure what exactly you need to know in terms of k-space. Again, I would very highly recommend Robin de Graaf’s book - he has an excellent chapter that should help you get a good first idea of k-space.

I’m not sure this is all immediately useful, but I think you have a lot of questions, so hopefully we can help you point towards the right resources over time.

Best,

Georg

1 Like

YesThese questions are related to the article you specified. But again in this paper though the operations before model fitting are called preprocessing(processing), operations after FT ( not model fitting) are called post processing…alittle confusing by this classification, an operation in frequency domain can be called both preprocessing and postprocessing…here was my problem…

and again in this article zero-filling is called a kind of spatial interpolation and the same time it is mentioned that it corresponds to interpolation of data in frequency domain

I consider spatial domain with image space and frequency domain with k-space…if Im right…with this definition of k-space in my mind that it is where spatial frequencies are stored after FT to each related to pixels of image…