Quantitative susceptibility mapping of the head-and-neck using SMURF fat-water imaging with chemical shift and relaxation rate corrections

MRM Papers Volume 87(3) - March 2022
1

link to paper

Quantitative susceptibility mapping of the head-and-neck using SMURF fat-water imaging with chemical shift and relaxation rate corrections

Beata Bachrata, Siegfried Trattnig, Simon Daniel Robinson

Abstract

Purpose

To address the challenges posed by fat-water chemical shift artifacts and relaxation rate discrepancies to quantitative susceptibility mapping (QSM) outside the brain, and to generate accurate susceptibility maps of the head-and-neck at 3 and 7 Tesla.

Methods

Simultaneous Multiple Resonance Frequency (SMURF) imaging was extended to 7 Tesla and used to acquire head-and-neck gradient echo images at both 3 and 7 Tesla. Separated fat and water images were corrected for Type 1 (displacement) and Type 2 (phase discrepancy) chemical shift artefacts, and for the bias resulting from differences in T1 and urn:x-wiley:07403194:media:mrm29069:mrm29069-math-0021 relaxation rates, recombined and used as the basis for QSM. A novel phase signal-based masking approach was used to generate head-and-neck masks.

Results

SMURF generated well-separated fat and water images of the head-and-neck. Corrections for chemical shift artefacts and relaxation rate differences removed overestimation of the susceptibility values, blurring in the susceptibility maps, and the disproportionate influence of fat in mixed voxels. The resulting susceptibility maps showed high correspondence between the paramagnetic areas and the locations of fatty tissues and the susceptibility estimates were similar to literature values. The proposed masking approach was shown to provide a simple means of generating head-and-neck masks.

Conclusion

Corrections for Type 1 and Type 2 chemical shift artefacts and for fat-water relaxation rate differences, mainly in T1, were shown to be required for accurate susceptibility mapping of fatty-body regions. SMURF made it possible to apply these corrections and generate high-quality susceptibility maps of the entire head-and-neck at both 3 and 7 Tesla.

2

Whether the susceptibility of fat is around 0.4 ppm is a big question. Here are the concerns:

  1. The authors used the PDF method to remove background phase. The PDF method in theory assigns susceptibility values to pixels outside the brain. As a result, background phase close to the surface of the brain, may be under- or over-corrected. When background phase is not completely removed, subsequent susceptibility measurements can be wrong.

  2. When we simulated phase due to the general geometry of an object, we found that we could not completely remove this phase within 10 pixels of the boundary, unless we know the exact boundary position to a fraction of a voxel. Please see Fig. 6 of https://doi.org/10.1016/j.mri.2018.07.009

  3. I was thinking whether water, fat, and iron can co-exist in one image voxel, especially in an iron overloaded liver. I would think so. In this scenario, one may find that fat has a noticeable susceptibility value but it does not.

3

Dear Yu-Chung Norman Cheng,

Thanks for your questions and a sincere apology for the (very) late response.

The points you raise - shortcomings in background field correction (BFC), problems generating susceptibility estimates close to the edge of the object and source separation - are enduring challenges in QSM - a field which is only gradually and incrementally earning the Q in its name. These don’t relate particularly to the issue addressed in this study, though. The topic we address is that of estimating susceptibility in tissues containing both fat and water. In that effort, we use quite standard QSM tools, selected from a large range of those available. Each of these is subject to its own strengths and shortcomings. We tested many of them for this application (e.g. for BFC, PDF, V-SHARP and others). Our conclusions are based upon assessments using those which were well suited to this application (judged by having low noise, low streaking, not overly smooth) but are not specific to them. By which I mean that other approaches would have (or did) yield similar results, but somewhat shrouded by artefacts.

On the difficulty of reliably estimating susceptibility at the edge of the object (with PDF, but also with other BFC techniques), SMURF imaging offers a partial solution in that it increases the signal-generating volume for which susceptibility can be estimated reliably, because correcting chemical shift artefacts leads to continuity of signal between fatty and non-fatty tissues where it would otherwise, i.e. with broadband excitation or fatsat, be interrupted by displacement of the fat signal or cancellation of the signal between fat and water. As such, there are fewer edges and boundaries in the images.

The final point you raise is the problem of estimating susceptibility in voxels containing tissues with different susceptibilities. The effect of these on B0 are additive, so that e.g. positive and negative sources cancel, leading to “incorrect” susceptibility values, or at least the inability to separate the contributing susceptibilities. Again this problem does not exclusively or particularly affect QSM of the head and neck (or regions containing fat), but also the brain, where the main sources are myelin, iron and calcium. This question is currently being addressed with methods which incorporate the effects that such mixed voxels have on the rate of signal dephasing, i.e. models which incorporate transverse relaxation rates (Chen et al., 2021, Shin et al., 2021),

Chen J, Gong NJ, Chaim KT, Otaduy MCG, Liu C. Decompose quantitative susceptibility mapping (QSM) to sub-voxel diamagnetic and paramagnetic components based on gradient-echo MRI data. Neuroimage. 2021 Nov 15;242:118477. doi: 10.1016/j.neuroimage.2021.118477. Epub 2021 Aug 14. PMID: 34403742; PMCID: PMC8720043.

Shin HG, Lee J, Yun YH, Yoo SH, Jang J, Oh SH, Nam Y, Jung S, Kim S, Fukunaga M, Kim W, Choi HJ, Lee J. χ-separation: Magnetic susceptibility source separation toward iron and myelin mapping in the brain. Neuroimage. 2021 Oct 15;240:118371. doi: 10.1016/j.neuroimage.2021.118371. Epub 2021 Jul 6. PMID: 34242783.

4

Dear Simon,

For whatever reason, I was not notified by your reply. I just see your reply today (March 28, 2023). Thus I went to check your papers again and refreshed my memory of the issues.

I am not sure that you noticed my core concern. Perhaps it is easier to state some facts first. As far as I remember, fat has a chemical shift but it has a very small susceptibility difference from the susceptibility of water. I just quickly search online and find this paper:
https://doi.org/10.1002/nbm.3766

Those authors measured a susceptibility of 0.1 ± 0.14 ppm for fat from liver. They also used the PDF method to remove the background phase. If they measured the susceptibility of fat from the middle of the liver, then that would support my first point above.

In addition, if fat has a susceptibility of 0.4 ppm as you have measured, then people should have definitely noticed that 30 years ago. This is because the susceptibility of venous blood is also about 0.4-0.45 ppm. Thus, a susceptibility of 0.4 ppm is not small. It should induce phase and people would have observed such phase before QSM was invented. But I am not aware that people had reported induced phase due to the susceptibility effect of fat in the body.

All these facts let me consider where the problems could be. It was why I provided my first two points.

Another issue you could have was to combine echoes for your phase analysis. I think you examined fat areas around the skull or in the neck. These areas are subject to the surface problem I mentioned above or low SNR due to nearby sinuses. In the latter case, when SNR is low at longer echo times, the uncertainty of phase increases. That will lead to a problem when you combined echoes.

I would like to mention several more points. It is a problem when anyone says “standard QSM tools.” I pointed out the fundamental math and physics issues in QSM in 2013 and 2014 QSM workshops. The partial volume effect is the major problem. Most papers do not include the partial volume effect in simulations. But our work do. Without the partial volume effect included in simulations, a method will always seem to work near perfectly of those simulated data. However that is far from the truth.

Susceptibility of each voxel cannot be accurately quantified from inverting phase maps (or field maps). This was studied by people who did SQUID research. See references

in:
http://dx.doi.org/10.1088/0031-9155/54/5/005

Our work mentioned in the above second point offered a way to properly remove the background phase. Yes, it is inconvenient but it does the job correctly. For phantom studies, one could unwarp phase first and then use SHARP, as we had shown in:
http://dx.doi.org/10.1002/mrm.26035
It turned out that combinations of different methods to remove background phase can also lead to different results. Thus choosing methods and their combinations is important too.

Separating different susceptibility sources has some problems too. I have gone through the two papers you mentioned as well as three papers by Emmerich et al. (Sina Straub’s group). All of their phantom results showed problems. I have also listened to the virtual workshop around the middle of March. At least during the virtual workshop, Jingjia Chen had correctly stated that this has to assume low concentrations of the sources. I would add, perhaps a few special cases. I will think more carefully on this topic and then I would like to write a commentary or something so people are not going to be confused in the future.

Thank you

Norman