Quantification of metabolites is one of the major challenges in current clinical MR spectroscopy. At higher field strength, the increase in SNR and spectral resolution helps to improve the reliability and reproducibili-ty of the quantitative results. The improvement has been striking for some metabolites, such as Glu, Gln and GABA, suggesting that multiplet resolution is as important in spectral quantification as SNR . Still, all the difficulties and pitfalls known from quantitative analysis of spectra acquired at 1.5 T, especially when trying to determine absolute concentrations, apply to spectra acquired at 3.0 T. The challenges of quantitative spectroscopy are twofold, as in a first step a variety of methods can be used to determine the integral under the individual peaks, which is known to be proportional to the concentrations, and then to translate the value of the integral thus determined into true concentration. While the first is more of a statistical and mathematical problem, albeit quite a large one, where all methods applied should yield similar results, the second step can adopt conceptually different approaches that can yield very different results. Apart from the statistical methods, there are two conceptually different techniques to approach a quantitative output MR spectroscopy:
1. The use of a reference as a standard to normalize the results
2. The use of the so-called reciprocity theorem to translate signal intensity directly into an absolute concentration value
There is no agreement on which method should be employed in which cases, but use of a reference has become an established method, at least as a significant set of results. There are several signals that can be used as a reference: historically, creatine has been used as a reference, as in the majority of the neurological diseases studied with 1H-MRS the Cr concentration has been shown to be constant up to the reliability of the method. On the other hand, a mounting body of findings where Cr is not constant is leading to the exploration of further reference markers. One of these markers is the internal water signal. Since the mmol concentration of water from healthy brain tissue is known, the metabolite signal can be normalized using the water signal as a reference. This method has disadvantages, as it does not take atrophy into consideration, even though there are methods to correct, for instance, for CSF contamination. If no internal signal is available, an external reference like a small phantom close to the sample volume has been suggested. This requires good B0 and also RF homogeneity, which is not always given. While the use of internal references works as well for 1.5 T and 3.0 T, the use of an external signal is more difficult. Due to the properties of wave propagation, it is nearly impossible to achieve homogeneous RF excitation over a large FOV with 3.0 T systems. It would be necessary to acquire the spectrum of the VOI inside the patient followed by an acquisition of an external reference while keeping all acquisition parameters constant.
This also yields to a problem related to the second conceptual approach using the reciprocity theorem to quantify in absolute units. The use of the theorem requires knowing the exact transmitter gain, e.g. defined by the RF energy needed to apply a 90° pulse, and receiver gains used to amplify and digitize the signal. Whereas this works well at 1.5 T, acquisitions at higher field strengths are impaired by dielectric resonance due to B1 inhomogeneities.
Additional to the analytical approach, the decrease in the intrinsic individual metabolic T1 and T2 relaxation times have to be well thought out , particularly if metabolite concentrations are to be expressed as absolute concentrations. The „normal" metabolite ratios observed at 1.5 T are, therefore, different at higher magnetic fields and it is essential to gather new normative data when switching to a different field strength.
Despite all the qualitative arguments speaking for MR systems with higher field strength, the quantification of the resulting spectra can be much more demanding, and some of the methods well established at 1.5 T might not even be applicable at all.
Was this article helpful?