MRS Quantitation

Can MRS peaks be quantified? How accurate is this?  

As a first approximation, the area under a given spectral peak should be directly proportional to the concentration of nuclei giving rise to that peak. It should be noted that this is not the same as metabolite concentration, as a single metabolite molecule may contain two or more NMR-active nuclei. 

Early attempts at clinical MRS quantitation followed methods used in chemistry laboratories — calculation of peak areas via simple numerical integration. It was soon realized that the peak areas depend on multiple additional factors, including the T1 and T2 values of each metabolite, Bo and B1 inhomogeneities, eddy currents, the type of pulse sequence, TR and TE. Even with these effects corrected, spectral line fitting and integration is still difficult because many MRS resonances overlap, are broadened by inhomogeneities, are superimposed on a wandering baseline, and have complex splitting patterns. 

In the 1990s accurate spectral curve fitting was significantly improved using advanced prior knowledge about the location and line shapes of expected metabolites. Such metabolite basis sets can be applied to extract a more quantifiable spectrum from the noisy raw data. This concept is applied to the most widely used algorithms today, Provencher's commercially available LCModel and several methods in the freeware jMRUI toolbox (available at www.jmrui.eu). 

Use of advanced prior knowledge to improve spectral line fitting. A basis set (left) incorporates chemical shifts, T2 relaxation times, and line shapes of expected metabolites. The basis set is used to extract a corrected spectrum of the detected components from noisy raw data (lower right). (From Boska et al, Mol Neurodegen 2014;9:58 under CC BY License)

Curve fitting programs can provide only relative metabolite concentrations, so additional methods or assumptions must be used to estimate absolute tissue concentrations from spectra. One possibility is to use an external reference standard (e.g. a test tube containing metabolites or chemicals at known concentrations) taped directly to the subject. This method introduces errors produced by B1 and Bo field inhomogeneities and susceptibility effects. It is more commonly used for non-¹H MRS.

Alternatively, MRS calibration can be made to an internal reference, such as tissue water, whose concentration is independently known. A commercial example of this is GE's PROBE-SVQ, which acquires a set non-water suppressed excitations as a reference prior to a conventional water-suppressed sequence. The concentration of brain metabolites can then be estimated by comparing their area ratios to that of tissue water, which is often assumed to be constant at 55.5 mol/kg.

The accuracy of these "absolute" measurements varies by technique and is in most cases unknown. Both internal and external calibration methods contain "correction factors" that warp measured values into reasonable physiological ranges. I am rather skeptical of any metabolite concentrations values reported to more than 2 significant digits, and my personal experience has been that even careful repeated measurements in the same patient may vary by 10-20%. I suggest that you follow the Elster Principle of healthy skepticism when assessing absolute values of metabolite concentrations at our current stage of technology.

Advanced Discussion (show/hide)»

Provencher's LCModel is commercially available and has been available since the mid-1990s. LC stands for “linear combination”, producing a solution representing the sum of contributions from individual metabolite spectra of pure compounds. The technique assumes spectral features such as chemical shifts, J-coupling, spectral distortions and localization errors will be identical across the sample. Exclusions may lead to a systemic bias.

The jMRUI toolbox contains several useful MRS quantitation and processing programs, including the relatively popular AMARES (Advanced Method for Accurate, Robust, and Efficient Spectral fitting). Unlike the LCModel which does fitting in the frequency domain, AMARES performs its estimation and corrections in the time domain.

Even sophisticated curve fitting programs like LCModel and AMARES have difficulty controlling for the undulating macromolecular baseline. A separate experiment may be required to measure this baseline where large differences in T1 relaxation between metabolites and other macromolecules can be used to null the metabolites of interest.

---

References
     Alger JR. Quantitative proton magnetic resonance spectroscopic imaging of the brain: a didactic review. Top Magn Reson Imaging 2010; 21:115-128. (good recent review)
     Barker PB, Soher BJ, Blackband SJ, et al. Quantitation of proton NMR spectra of the human brain using tissue water as an internal concentration reference. NMR Biomed 1993; 6:89-94. 
     Gasparovic C, Song T, Devier D, et al. Use of tissue water as a concentration reference for proton spectroscopic imaging. Magn Reson Med 22006; 55:1219-1226.
     Jansen JFA, Backes WH, Nicolay K, Kooi ME. ¹H MR spectroscopy of the brain: absolute quantification of metabolites. Radiology 2006; 240:318-322.     
     Mandal PK. In vivo proton magnetic resonance spectroscopic signal processing for the absolute quantitation of brain metabolites. Eur J Radiol 2012; 81:e653-e664. *(good review)   
     Poullet J-B, Sima DM, Van Huffel S. MRS signal quantitation: a review of time- and frequency-domain methods. J Magn Reson 2008; 195:134-144.
     Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993;30:672-679.
     Provencher S. LCModel & LCMgui User's Manual. 2016. (Complete documentation of most current software version available from this link)
     Stefan D, Di Cesare F, Andrasescu A, et al. Quantitation of magnetic resonance spectroscopy signals: the jMRUI software package. Meas Sci Technol 2009; 20:1040355 (9 pp).
     Vanhamme L, van den Boogaart A, Van Huffel S. Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J Magn Reson 1997; 129:35-43. (Description of AMARES method).

---

Related Questions
     What causes magnetism?

---

←  Previous Question

Next Question  →

↑ Complete List of Questions ↑