is a Radiology Resident in the Department of Radiology at Tripler
Army Medical Center, Honolulu, HI.
is a Neuroradiologist at Tripler Army Medical Center, HI; and an
Assistant Professor in the Department of Radiology, Uniformed
Services University, Bethesda, MD.
Portions of this article were previously presented as a
poster presentation at the ASNR annual meeting in San Diego, May
1999; as well as in a presentation, MRS: The Good, The Bad, and
The Ugly, at Ski the Sky, Big Sky Radiology Course, Big Sky, MT,
Magnetic resonance spectroscopy (MRS) is becoming more
frequently used and easier to perform. With a software package
addition, an MR instrument can be fully equipped to perform
multiple types of MRS. Faster slew rates and gradients provide
improved signal-to-noise ratios, even when a smaller voxel sample
size is selected.
This article will attempt to simplify the basic MRS concepts in the
context of data acquisition and interpretation employed in
neuroradiology. The concepts presented here are based on
single-voxel MRS with a repetition time (TR) of 1700 msec and an
echo time (TE) of 30 msec. Even so, the basic concepts are the same
with respect to choline, lactate, lipid, and N-acetyl aspartate.
These concepts can be utilized with two-dimensional (2D) or
three-dimensional (3D) imaging, using smaller voxel size or
variations in voxel size.
MR spectroscopy is a complex field employing the basic
principles of nuclear magnetic resonance spectroscopy (NMR) and MRI
to obtain clinically relevant information. This article will not
attempt to explore the more subtle nuances of MRS nor discuss the
numerous qualitative measurements that can be performed. The goal
of this article is to allow the general radiologist to feel
comfortable with the basic principles of reading MRS and to be able
to decide rapidly whether the pattern is abnormal or normal. We
will briefly review several types of MRS that radiologists can
perform, the three main metabolites, some theory on why MRS changes
with various pathologic processes, and how to identify and
interpret some main patterns of the spectral graphs.
What is MR spectroscopy?
Rather than displaying MRI proton signals on a gray scale as an
image depending on its relative signal strength, MRS displays the
quantities as a spectrum. In routine MR imaging, the more edema on
a T2 sequence from mobile protons, the brighter the signal on T2.
Using MRS, the more metabolite that is present, the taller the peak
or greater the area under the peak. Specific metabolites can be
located along an x-axis, which is expressed in parts per million
(ppm), a dimensionless unit. We can infer that a spectrum of the
brain is normal from numerous years of studying such spectra in
healthy subjects in whom peak positions and relative intensity
ratios have been established (Table 1 and Figure 1).
Once an MR image is obtained as a localizer image, a volume of
interest is selected. If a single voxel is to be analyzed, then a
single 3D region of interest is selected. Once the single voxel is
obtained, the spectrum is collected based on the amount of protons
in the voxel. The proton signals are detected and represented as a
free induction decay (FID). A Fourier transform is applied to the
FID, converting the temporal information into frequency
information. The resonant frequency is then plotted versus signal
intensity on a spectrum, instead of the typical gray-scale image.
If multiple voxels are to be evaluated, then both a region of
interest for evaluation and a region of normal brain are selected
for comparison (Figure 2). Of the single-voxel techniques, two
commonly used acquisition sequences are stimulated echo acquisition
mode (STEAM) and point resolved spectroscopy (PRESS). With twice
the signal of STEAM, PRESS acquisitions are faster; however,
spectral baselines are better with STEAM sequences. Care must be
taken when identifying voxels of interest (especially for the
normal brain comparisons), since significant regional differences
in metabolite distributions can be seen in both gray and white
matter (Figure 3).
Regions to be avoided when selecting voxels include blood, bone,
and cysts, since susceptibility artifacts may skew the expected
normal molecular distributions. Areas that are difficult to image
include the posterior fossa and the spinal cord (both of which
encounter problems due to their proximity to bone), as well as
tumors containing cystic components, blood, or regions of
Main metabolitesThe Good, The Bad, and The
Although there are several metabolites included in the spectrum
of single-voxel MRS, we will review the most important in this
There are three main players in MRS of the brain: N-acetyl
aspartate (NAA), choline (Cho), and the lactate and lipid groups
(LL) (Figure 4). An explanation of these three metabolites can
bring most physicians up to speed on the basics. Although there are
several theories as to why the relative concentrations of each
this introduction will briefly discuss some basic theories behind
why each metabolite may change from the normal ratio.
NAA = neuronal health (The Good)
N-acetyl aspartate is seen at 2.02 ppm and is believed to be a
marker of neuronal health. Originally, decreases in NAA were
considered to be due to neuronal destruction, since it was
diminished in cases of multiple sclerosis and following trauma.
Since it can be reversible, however, this is probably better
described as a marker proportional to the health of the neurons.
Higher peaks indicate more normal neuronal presence, while
diminished peaks occur in situations in which neural damage or
replacement has occurred.
Cho = Tumor marker or cell wall marker (The Bad)
Choline is seen at 3.22 ppm and is present in cell walls of
normal brain tissue. As more brain cells are made, one theory
suggests the Cho is increased. Active tumor growth will then cause
an increase in Cho, since there is above-normal production of
cells. Other processes can release or increase Cho besides tumor;
multiple sclerosis or acute infarctions will also release Cho, or
cause lysis of cell walls, and increase the concentration of Cho.
This can be a transient effect, however, while tumors will
demonstrate persistent Cho elevation. We call it The Bad, since
tumors show an increase in this metabolite.
LL = Destruction and necrosis (The Ugly)
Lactate (lactic acid) is seen as a doublet (two peaks close to
one another) at 1.33 ppm and is a by-product of anaerobic
metabolism. Lipids resonate at the 0.9 to 1.2 ppm range. Both are
released with cell destruction or synthesized in necrosis.
Increased LL can be seen in necrotic tumors, and in stroke due to
destruction of cells, and in abscess. It can also be seen in lower
concentrations in intermediate tumors. Lactate and lipid peaks are
generally present in aggressive disease processes.
The three-step approach to spectral analysis
The quality assurance phase. Is it an adequate spectrum?
Is Hunter's angle normal?
Starting from the right side of the graph, count off the location
and check quantities of The Good, The Bad, and The Ugly. These are
located on the x-axis at 2.02 ppm, 3.22 ppm, and the area from 0.9
to 1.33 ppm.
Just as a bad image can make interpretation difficult or
impossible for diagnosis, a bad MRS may not be interpretable.
Substances that are difficult for MRS to image include bone, blood,
cysts, and cerebral spinal fluid (CSF). It is difficult to obtain
spectra of bone and blood due to immobile protons (bone) and shim
difficulties (blood). Both CSF and cysts can contain lactate
products and, thus, may lead to inaccurately elevated lactate or
lipids as well. When performing voxel measurements, you should stay
clear of these substances in all three imaging planes. Since the
area sampled is a voxel, it acquires signal from regions above and
below the box that has been placed.
Examples of an adequate spectrum include good water suppression;
otherwise the water peak on the MRS spectrum is so abundant, it
will overshadow the other metabolites. The peak occurs at the far
left of the spectrum at 4.7 ppm. (Figure 5). Without suppressing
the water signal, it will overwhelm signal from the other
metabolites and result in an inadequate spectrum where evaluation
of peaks with smaller contributions becomes more difficult.
Hunter's angle is a term coined from a neurosurgeon, Hunter
Sheldon, at Huntington Medical Research Institutes. Instead of
doing complex ratios and analysis of the spectra, he simply used
his pocket comb. He placed his comb on the spectrum at
approximately a 45š angle and connected several of the peaks. If
the angle and peaks roughly corresponded to the 45š angle, the
curve was probably normal (Figure 6). If the peaks strayed off the
comb's angle, the curve was abnormal (Figure 7). This is a quick,
useful method to read MRS and determine normal from abnormal. It is
important to remember, however, that this angle was used with STEAM
spectra from the brain. This article will not address normal
spectra elsewhere in the body.
The Good, The Bad, and The Ugly
We can look at NAA, Cho, and LL in a more simplified pattern
(Figure 4). First, the zero point on the curve is located at the
far right of the x-axis in spectral analysis. N-acetyl aspartate is
a neuronal marker, thus making it a high, plentiful peak on the
curve in normal brain tissue. We can call this The Good marker. If
the neuronal health is good, this peak will be the highest peak. It
is located at 2.02 ppm on the x-axis. Elevations do not occur
(except in patients with Canavans disease). Decreased NAA can be
due to replacement with other metabolites (ie, tumor cell walls) or
due to unhealthy neurons, as in diffuse axonal injury, multiple
sclerosis, or infarction. It can be reversible.
Next, we moved left on the spectrum to 3.22 ppm on the x-axis,
the location of the Cho peak. Remember, we termed Cho The Bad
because excess amounts can indicate cell destruction and release of
cell walls, or an increase in cell wall synthesis. Excess Cho is an
indicator of tumor. It can also be elevated in early phases of
cellular destruction and lysis, as in multiple sclerosis and
stroke. These can therefore mimic tumor in their early phases.
Finally, there are the LL peaks located between 0.9 and 1.33
ppm. This is termed The Ugly because it is an extremely dreadful
finding. Lactate and lipid peaks occur when necrosis and a sizeable
amount of cell death occurs. It will be the highest peak on the
spectrum in most high-grade tumors with marked depression of the
NAA peak. Another cause of increased LL peaks occurs with cellular
destruction such as stroke.
Predominantly The Bad and The Ugly. There is abundant LL and Cho.
N-acetyl aspartate is depressed from replacement of neurons with
cell wall synthesis and necrosis (Figure 8).
Predominantly The Bad. There is elevated Cho from tumor cell wall
synthesis, but not marked elevation in LL from necrosis. Some NAA
depression is present (Figure 9).
Stroke or radiation necrosis:
Predominately The Ugly. There are decreased NAA and Cho peaks
with elevation of LL from destruction (Figure 10).
Loss of The Good. There is loss of NAA peak height, but not much
elevation in Cho or LL chronically. Early on, both Cho and LL can
be elevated (Figure 11) and can mimic tumor. A follow-up MRS will
usually demonstrate change.
These are the basic concepts in reading MRS. You may think of
MRS as an additional MR sequence. It may help to focus on a
differential diagnosis and can be used in conjunction with other
imaging. Though there are many more innuendoes that can be learned
along the way, remember to systematically ask the three
1) Is it an adequate spectrum?
2) Is Hunter's angle normal or abnormal?
3) How would The Good, The Bad, and The Ugly be measured?
Starting from the right, count off and locate heights at the 2.02
ppm, 3.22 ppm, and 0.91.33 ppm areas.
Not enough Good--Trauma or multiple sclerosis.
Too much Bad--Usually tumor.
Too much Ugly--Stroke or radiation necrosis.
Too much Bad and Ugly--High-grade tumor.
MR spectroscopy can help to limit a differential diagnosis. It
provides a look at what cellular products may be present in the
brain. MR Spectroscopy can be used to answer specific clinical
questions, such as differentiating between radiation necrosis and
tumor recurrence (Figure 12) or to determine where the most
aggressive portion of the tumor is located for biopsy.