In-Vivo Magnetic Resonance Spectroscopy I: Probeheads and Radiofrequency Pulses Spectrum Analysis

NMR experiments are designed to selectively perturb an ensemble of nuclear spins that have been aligned to a strong static magnetic field B ₀ using a second radiofrequency (rf) field B ₁, and to monitor the specific electromagnetic message sent out by these spins upon their return to equilibrium. The rf probe has a dual role in this game: it delivers the perturbation energy to the spin system, and it receives the rather weak response signal from the spin system under test. Desirable properties of this rf “interface” include good power efficiency for the excitation, and high sensitivity for signal reception, with an homogeneous distribution of the field and reception sensitivity across the whole sample space.

The growing interest in the application of NMR spectroscopy to intact biological systems has placed unique demands on instrumentation. Perhaps the most critical piece of instrumentation for the in-vivo spectroscopist is the NMR probe. The need to perform quick and reliable experiments, particularly in the clinical setting, makes probe sensitivity critical. Many studies involve the use of more than one nucleus. Typically, proton NMR is used to shim, obtain localizing images and/or spectra. Additional spectra may be acquired from ³¹P, ²³Na, ¹³C or other nuclei. This multinuclear approach to in-vivo NMR spectroscopy has placed a premium on multiple tuned NMR probes that perform with high sensitivity. There are many designs for double tuned probes in the NMR literature. Some designs for probes tuned to three and four nuclei have also been presented. All of these designs must sacrifice some sensitivity at one or more of the resonant frequencies at which they operate when compared to a similar single tuned coil. One common strategy of all of the multiple tuned probe designs is to minimize the loss in sensitivity. In most designs it is possible to arbitrarily distribute the loss in sensitivity between the different operating frequencies of the probe by proper choice of component values. Thus the evaluation of any multiple tuned probe can only be performed if the probe is optimized for a particular application. A second, less frequently discussed aspect of multiple tuned NMR probes is orthogonal tuning adjustment. In the in vivo setting it is very difficult to control loading of the probe. Therefore, the tuning must be adjusted with each application. Independent tuning of the individual resonances is important to ensure reliable performance by the probe. In this paper, I will discuss in detail many of the multiple tuned designs that have appeared in the NMR literature, paying particular attention to the advantages and disadvantages of each. I will focus on multiple tuned surface coils, but will also treat double tuned volume coils, which are of importance in the context of volume selective MRS techniques.

Despite its obvious potential for assessing in vivo metabolism, NMR spectroscopy has yet to acquite a significant place in clinical practice. This situation can be partly ascribed to the technique’s inherently low sensitivity and to the low metabolite concentrations. Thus, compared to the microliter resolution of clinical MRI, metabolite spectroscopy suffers from low spatial resolution. When acquiring larger volumes (1–100 ml) to attain sufficient signal, specificity may be lost due to tissue heterogeneity. This greatly complicates the reproducibility of clinical trials, especially in cases where the size (or position) of the lesion changes. For the purposes of sensitivity and access to many metabolites, it is advantageous to use proton NMR. However, proton spectroscopy is hampered by its limited spectral resolution and by the need to effectively suppress the dominating water and lipid resonances. The strategies that can be employed to attain suppression of these and other unwanted resonances are the topic of this review.

Inhomogeneities in the B ₁ field pattern of RF transmitters are often the cause of major problems in applications of the NMR phenomenon. This simply follows from the fact that NMR experiments in general rely on rotation of magnetization vectors about the B ₁ field through an angle γ ∫ B ₁(t)dt, where γ is the gyromagnetic ratio. Consequently, even very simple tasks such as spin-excitation becomes non-uniform over the sample and this deleterious effect gets compounded in more complex and demanding pulse sequences employed in contemporary NMR applications ranging from high resolution studies of molecular structure to imaging and localized spectroscopy in vivo. This problem is particularly pronounced in the human and intact animal investigations in vivo where geometrical constraints frequently require the use of transmitters and detectors, such as surface coils, which generate extremely non-uniform B ₁ field patterns.

Selective pulses were originally proposed by Garroway, Grannell and Mansfield [1] and have come to play a central role in the majority of magnetic resonance imaging (MRI) experiments, where they are used in conjunction with an applied linear magnetic field gradient to isolate a slice of controlled thickness and orientation from within an extended object [2, 3]. Selective pulses feature too in many of the more successful magnetic resonance spectroscopy (MRS) methods [3–5]. Although at present little exploited, they also hold great promise for conventional high resolution spectroscopy. Potential applications include solvent suppression without distortion of the frequency response, magnetization transfer and reduction of bandwidth in two- and three-dimensional high resolution experiments.

One of the most important techniques in biological and medical Magnetic Resonance (MR) is the spatially selective radiofrequency (rf) irradiation. With this tool the MR signals originating from a large heterogeneous sample like the human body can be focussed to a small area of interest to perform refined chemical and anatomical investigations. Nowadays the technique has a broad range of applications: it defines an imaging-slice in two-dimensional (2D) MR imaging, it selects volumes for 3D imaging, it predefines regions for spectroscopic imaging, it saturates areas to avoid flow and motion artefacts and it prepares volumes for localized spectroscopy.

Although a great deal of useful information is obtained from relative concentrations, absolute concentrations can further enhance the usefulness of in vivo NMR spectroscopy (MRS). Because of the non-invasive nature of MRS, clinical measurements of absolute concentrations can be made (and repeated if required) on tissues such as neonatal brain that are normally inaccessible to biopsy. Furthermore, because observations can be made on tissue in situ, metabolite concentrations can be obtained without the risk of the freezing artefacts that can affect biopsy results. With the absence of reliable direct methods of absolute quantitation, early attempts to estimate metabolite concentrations in human tissue by MRS (mainly in muscle [1–5] and brain [6–10]) were indirect and depended on knowledge of either the total mobile phosphate (P_tot) concentration or the concentration of a particular, normally stable metabolite (e.g. adenosine triphosphate (ATP)). For instance, in studies of normal, resting, human muscle [1, 3] phosphocreatine (PCr) and inorganic orthophosphate (Pi) concentrations have been estimated from relative peak areas assuming either [ATP] = 5.5 mmol kg⁻¹ wet wt or [P_tot] = 49.5 mmol kg⁻¹ wet wt (see Table 1). In calculations based on [ATP], absolute concentrations were derived using the β-ATP resonance because, unlike the γ and α resonances, the β peak does not have contributions from other underlying resonances. (The γ and α peaks include contributions from adenosine diphosphate (ADP) and the α peak also includes contributions from nicotinamide dinucleotides (NAD + NADH).) In the calculations based on [P_tot] the following was assumed: These two approaches gave very similar results and favourable comparison with biopsy data. In fact, due to the inadequate cessation of metabolic processes prior to and during the freezing of biopsy samples, it appears that MRS has the potential for providing better measurements of [Pi] and [PCr] than biopsy. However, in many important situations these simple approaches are inadequate because of the invalidity of the assumption of stability for the concentration reference and therefore alternative methods are required.