The voltage detected by the spectrometer from a sample containing NAA can be expressed as:

Snaa = P x [NAA] x v x f(T:NAA, T2NAA, TR, TE, Bi) [1]

where v is the MRS voxel size (in cm3), [NAA] is the millimolar concentration of NAA, f() is a "pulse sequence modulation" factor that will depend on the pulse sequence used, its repetition and echo times (TR and TE), the T1 and T2 relaxation times of NAA, and B1 is the strength of the radiofrequency field within the MRS voxel. P is a scaling factor that can be expressed as:

where NS is the number of scans (averages) performed, G is the receiver gain, ro0 is the spectrometer operating frequency (e.g. 64 MHz for a 1.5 Tesla magnet) and Q, n, ^,Vc, and a are all factors related to the geometry and quality of the radiofrequency receiver coil (5). The proportionality constant for Eq. [2] is unknown, prohibiting the calculation of [NAA] without additional calibration measurements. However, Eqs. [1] and [2] do indicate how the signal (and hence sensitivity) of NAA can be enhanced, by using large numbers of scans, larger voxel size, high field systems, sensitive RF coils (high B1), and maximizing the modulation factor f (by choosing optimum TR and flip angles, and short TE). Sensitivity is usually largely independent of the receiver gain, G, since this usually increases both signal and noise equally.

In order to calculate NAA concentrations, a reference signal must be acquired, ideally under identical conditions to those used to record the NAA signal, so that all proportionality factors in Eqs. [1] and [2] are identical, allowing the ratio equation to be written:


Sreference where we assume from hereon that Snaa and Sreference have already been corrected for possible differences in T1 and T2 relation times according to standard equations (6).

The choice of the reference compound is of key importance for the accuracy of the quantitation procedure. The most commonly used method is to select some other compound in the brain spectrum (e.g. most often creatine (Cr)) and report ratios of NAA/Cr. This is an example of an internal intensity reference, namely one that comes from the same voxel as the signal to be measured (i.e. NAA). Internal references have the advantage that many of the factors in Eqs. [1] and [2] are virtually identical for both signals (e.g. volume of tissue, B1 (and B0) field strength, flip angle and other pulse sequence related factors), and so are insensitive to systematic errors associated with these parameters. Using a reference signal from the same spectrum (as the compound to be measured) also has the advantage that no additional scan time is required. However, while initially it was hoped that Cr levels might be relatively constant throughout the brain and invariant with pathology, subsequent studies have shown both substantial regional (6,7) and pathology-related changes in Cr (8), so that in general it is somewhat unsafe to infer changes in NAA from the measurement of only the NAA/Cr ratio. The same comment also applies to other compounds detected in the brain spectrum (e.g. choline (Cho)) that might be considered as potential reference signals. An alternative, and widely used, internal intensity reference is the unsuppressed tissue water signal (9-11), which can be easily and quickly (at least for single voxel spectroscopy) recorded by turning off the sequence water suppression pulses. Brain water content is relatively well known, and pathology-associated changes are relatively small. Furthermore, it is possible to estimate voxel water content from appropriate MRI sequences (12). Finally, the unsuppressed water signal may also be helpful for phase- and eddy current-correction of the water-suppressed spectrum (13,14). For all these reasons, quantitation of single-voxel spectra using the internal tissue water signal has become a popular technique over the last few years. Studies have demonstrated that this is a reliable method, e.g. for multisite trials of brain spectroscopy (15).

However, there may be situations where water referencing is not optimal, for instance where brain water content is variable or not well known (for instance in neonatal studies, or pathologies involving major changes in brain water content). Also, water referencing may not be convenient in certain MR spectroscopic imaging (MRSI) studies. For instance, it may be prohibitively time consuming to record both water suppressed and non-suppressed MRSI datasets. In these instances, other approaches to quantitation should be considered. These approaches can be considered to fall into two classes, either external references, or the phantom replacement technique. External references involve the recording of a reference signal from a region outside of the primary region of interest. Often, a vial of known concentration compound (or simply water) is placed next to the head, and the signal from this region measured, either before or after the brain spectrum is measured. While external standards have the advantage that the reference concentration is exactly known, their use is complicated by the fact that some of the factors in Eq. [1] may no longer be constant. The biggest source of error is likely differences in B1 field strength (and probably also RF pulse flip angles) due to inhomogeneities in the RF coils used for reception and/or transmission. The external standard may also induce magnetic susceptibility effects that degrade the B0 field homogeneity, again leading to systematic errors because of suboptimal spectral quality. Also care has to be taken to ensure that the spatial localization sequence interrogates the same volume of tissue and reference sample. Sometimes a spectrum from a different (e.g.

contralateral) brain region can be used as a reference, e.g. in patients with focal brain disease, if the contralateral hemisphere is known to be normal. This approach avoids the susceptibility problems associated with the placement of an external vial, but does share the other potential problems of external references. Also, in many patients, it may be unsafe to assume that metabolite concentrations in apparently uninvolved brain regions (i.e. with normal brain MRI) are in fact the same as in the normal populations (16).

The phantom replacement technique can be regarded as a hybrid of external and internal intensity references. The basic idea is to record a spectrum from a patient, remove the patient from the magnet, insert a standard sample, and then record its spectrum using as closely matched experimental conditions as possible. The method has the advantage that the reference concentration is exactly known, and most of the factors in Eqs. [1] and [2] are also known and can be controlled for. One major difference between the two acquisitions, however, is that the RF coil "quality" factor (Q) will almost certainly be different, because the electrical properties (impedance) of the phantom will be different from the human head. Fortunately, this loading factor (F) can be readily determined from the RF power (voltage) calibration required to obtain a 90° pulse, and applied as a correction factor:



It may also sometimes be necessary to include correction factors if the brain and reference scans are collected with different numbers of scans and/or different receiver gains. The phantom replacement technique is convenient to use for MRSI scans, since it does not require any additional patient scan time, and, in fact, on stable clinical scanners, the reference scan may change little from one day to the next, so that it does not need to be recorded for every patient. Potential errors can occur due to B1 inhomogeneity if the brain and reference scans are from different locations. The method therefore works best with highly homogeneous transmit and/or receive coils (such as quadrature birdcage head coils found on most 1.5 Tesla scanners), otherwise care has to be taken to attempt to match the reference scan location as close as possible to those in the brain. Alternatively, an approach (described below) for correcting MRSI scans for B1 inhomogeneity prior to quantification (e.g. for use with phased-array receiver coils) can be used.

Table 1 contains a summary of the different quantitation methods available for MR spectroscopy and spectroscopic imaging, and lists their relative advantages and disadvantages.

Table 1. Summary of relative advantages and disadvantages of commonly used quantitation techniques for proton MR spectroscopy of the human brain.


Table 1. Summary of relative advantages and disadvantages of commonly used quantitation techniques for proton MR spectroscopy of the human brain.



Simple, no extra scan time, no B0 or B! errors

Pathological and regional variations common


Minimal extra scan time, simple, no B0 or B! errors

May change up to ~20% in pathological conditions

0 0

Post a comment