Dr. Ballestero is a diagnostic radiology resident at the University of Pittsburgh Medical Center, Pittsburgh, PA. She received her MD in 1997 from the Medical College of Pennsylvania.

In 1973, nuclear magnetic resonance was first applied to the field of medicine and, circa 1979, the first magnetic resonance (MR) image was created on human pathology. ¹ While magnetic resonance imaging (MRI) rapidly became an integral part of diagnostic imaging, magnetic resonance spectroscopy ( ¹ H-MRS) lagged behind. However, since its 1995 Food and Drug Administration (FDA) approval for neuroimaging, ¹ H-MRS has flourished. The metabolic and biochemical information that MRS provides, as well as its relative ease of acquisition and interpretation, have made it a useful addition to the MRI sequences that are now standard in all diagnostic radiology practices.

For many years, single-photon emission tomography (SPECT) and positron emission tomography (PET) have been used to identify metabolic changes involved in pathologic processes of the brain. However, the radiation exposures inherent to these techniques have limited their use. Many neurologic disorders are chronic in nature and require serial imaging over long periods of time. ¹ H-MRS can provide the same, if not more accurate, metabolic information without the ionizing radiation associated with SPECT and PET imaging. The ability to correlate anatomic and physiologic information by coupling ¹ H-MRS with MRI is an added advantage of this technique.

This article is intended to introduce the reader to some of the most basic principles of ¹ H-MRS and the most common clinically significant metabolites that are essential to understanding and interpreting a human ¹ H-MR spectrum. In addition, several cases will be presented to demonstrate the clinical application of ¹ H-MRS with respect to space-occupying lesions of the brain.

The fundamentals of ¹ H-MRS

Volume of interest --In order to produce a ¹ H-MR spectrum, a single MR image is first acquired to be used as a localizer. From this image, a volume of interest (VOI) is isolated. If single-voxel ¹ H-MRS is to be performed, the selected VOI will ideally surround only the tissue to be examined. For multi-voxel ¹ H-MRS (see below), a larger volume is selected so that the spectra from adjacent, more normal tissue can be simultaneously acquired and compared to the anatomic region of interest. As with all MR imaging, the larger the VOI, the greater the signal-to-noise ratio (SNR) of the resultant spectrum. Similarly, signal averaging is frequently invoked to increase the SNR to clinically usable levels. ^1,2

Localization --Three orthogonal slices are excited within the same pulse sequence. Their common intersection defines the VOI from which the ¹ H-MR spectrum will be obtained. This would represent single-voxel ¹ H-MRS. Should the study be repeated numerous times with phase encoding performed along two orthogonal axes, it would permit spatial localization within the VOI itself. This technique would be referred to as multi-voxel ¹ H-MRS. The latter has the advantage of increased SNR and superior in-plane spatial resolution of spectra within the targeted voxel. However, it is accompanied by a scan time that is increased proportional to the number of phase encoding steps performed.

In single-voxel ¹ H-MRS, the most commonly used techniques are the stimulated echo acquisition mode (STEAM) and point resolved spectroscopy (PRESS). STEAM has, historically, had the advantage of requiring shorter echo time (TE) studies (e.g., 30 ms), a factor necessary to detect the clinically significant neurometabolites with a higher SNR. PRESS, on the other hand, has the predominant advantage of producing greater spectral SNR (two-fold greater than with STEAM). Rapidly developing ¹ H-MRS technology has allowed the use of shorter TEs with the PRESS technique. As a result, PRESS has become the ¹ H-MRS method of choice in clinical use. ^1,3

Repetition time and echo time --MRI and ¹ H-MRS use similar techniques for data acquisition but present the data quite differently. MRI maps signal intensities to spatial localization data. The result is an MR image. ¹ H-MRS maps signal intensities to their component frequencies, creating the ¹ H-MR spectrum, or frequency map, of the examined tissue. It therefore stands to reason that just as variations of repetition time (TR) and TE will change signal intensities in an MR image, they similarly affect signal intensities (i.e., spectral peak heights), in a ¹ H-MR spectrum (figure 1). The shorter the TR, the less the longitudinal magnetization recovery of each metabolites' magnetization and, therefore, the weaker the resultant signal (i.e., smaller spectral peak heights). Similarly, the longer the TE, the more transverse magnetization of each metabolite will have decayed and, thus, the weaker the signal obtained, again resulting in smaller and fewer peaks. TR and TE must be kept in mind when designing a ¹ H-MR study and interpreting a ¹ H-MR spectrum. Valuable information can be obtained and/or lost depending on the selection of such timing parameters. ^1,3,4

Water suppression --The concentrations of the clinically significant metabolites can be 10,000 to 100,000 times smaller than that of free water. As a result, the area-under-the-water portion of the ¹ H-MR spectrum (at 4.7 ppm) tremendously overshadows the spectral peaks of the metabolites of interest (figure 2). Therefore, water suppression techniques are routinely employed in an attempt to eliminate the signal from free water.

The most commonly employed water suppression technique is the chemical shift selective (CHESS) radiofrequency (RF) pulse. This works in much the same way as the water and fat suppression pulses used in MRI. Prior to initiating a ¹ H-MRS sequence, the CHESS method will place a narrow, frequency selective RF pulse at the water resonance frequency, resulting in saturation of free water. When performed properly, this can lower the resultant water peak by up to 1,000 times, enabling the smaller metabolite peaks to be successfully detected and differentiated. ^1,5

Magnetic field homogeneity --Within a selected VOI, heterogeneities in the magnetic field will result in broadening and lowering of the individual component peaks. This results in poor spectral resolution (i.e., decreased ability to detect and differentiate individual peaks) while simultaneously worsening the SNR of the study. As in MRI, shim coils are used to correct for magnetic field distortions to yield narrow line widths and improved SNR; all in clinically feasible scan times. ^1,3

¹ H-MRS of the normal brain

MRS is used to identify and quantify metabolites in and around the tissues being examined. In order to detect pathology and differentiate it successfully from normal tissue, one must understand how a ¹ H-MR spectrum is read and be able to recognize normal spectral patterns. The roles of the clinically significant metabolites must also be addressed, as changes in their concentrations will provide biochemical/metabolic information about the tissue in question. With this knowledge, ¹ H-MRS can be used to develop or narrow a differential diagnosis of pathologic and/or normal entities.

The basic ¹ H-MR spectrum --A ¹ H-MR spectrum is essentially a graphic representation of the quantity and variety of metabolites present in a particular volume of tissue (figure 3). Each metabolite is plotted as a total quantity, mapped to its particular resonant frequency. The vertical axis is used to measure concentration. In simplest terms, the area under a peak represents a given metabolite's concentration. The greater the concentration, the taller and/or wider the peak. Poor magnetic field homogeneities will result in a lower peak height. Because the metabolite concentration is constant, the area under the peak will be spread over a broader range, thus resulting in lowered spectral resolution (the broader the peak, the greater the chance of overlap with other peaks). The horizontal axis is referred to as the frequency, or chemical shift, axis. It defines characteristic resonant frequencies of the detected metabolic moieties. The metabolite peak positions (typically reported as chemical shifts of parts per million (ppm) will not vary from subject to subject, as long as the same primary magnetic field is used (Table 1). ¹

The metabolites --As there are hundreds of metabolites produced by the human brain, the concept of mapping these out, using ¹ H-MRS, can be very daunting. However, because only a handful of these occur in significant quantities in both healthy and diseased states, most of the neurometabolites can be eliminated, making the interpretation of ¹ H-MRS a much less complex task. The clinically significant neurometabolites that one may encounter in clinical practice are as follows: ^1,6

N-Acetylaspartate (NAA): Found solely in the brain, NAA is a measure of neural tissue viability. As it is present in both gray and white matter, it is considered a marker of neuronal and axonal integrity. NAA creates what is typically the tallest peak in the normal brain spectrum. Because its decrease or disappearance can be due to either cell death or axonal injury, correlation with the patient's clinical history is essential. Decreased NAA can be seen in numerous pathologic processes including neoplasm, infection, ischemic injury, and demyelinating diseases. Elevation of the NAA peak is rare but may be seen in hyperosmolar states and axonal recovery.

Choline (Cho): Phosphotidylcholine, the most common choline neurometabolite, is found in lipid-containing structures such as myelin and cell membranes. When pathologic processes alter or destroy these structures, phosphotidylcholine is released and, thus, elevations of the choline peak will occur. Some of these processes include neoplasm, ischemic injury, and acute demyelinating diseases. Decreased choline peaks can also be seen, raising suspicion for infectious processes.

Creatine (Cr): In normal brain metabolism, phosphocreatine supplies phosphate to adenosine diphosphate (ADP), resulting in the production of an adenosine triphosphate (ATP) molecule and the release of creatine. Thus, the total creatine (creatine plus phosphocreatine) is a very reliable marker of brain metabolism. Creatine is generally seen as the constant, central peak in a ¹ H-MR spectrum. Lower creatine peaks may be seen in primary cerebral lesions such as neoplasm, ischemic injury, and infection. However, decreased creatine peaks may also be encountered in systemic diseases. Creatine, phosphocreatine, and their main precursor, guanidinoacetate, are synthesized in extracerebral tissues (primarily liver and kidney) and then transported to the brain. Therefore, any metabolic defect resulting in decreased production of creatine/phosphocreatine (e.g., he-patic or renal diseases) will lower the total creatine peak. Absence of the creatine peak can also be seen. Most neoplastic tissues that metastasize to the brain do not possess creatine kinase, and, thus, cannot produce creatine. When a ¹ H-MR spectrum lacks a creatine peak, metastatic tumor should be suspected.

Lactate (LAC): When oxygen supplies are depleted, the brain responds by switching to anaerobic glycolysis. One end product of this metabolic pathway is lactate production. Therefore, the presence of a lactate peak in a ¹ H-MR spectrum is a sign of hypoxic tissue. The etiology of decreased oxygen supply may include: decreased oxygen delivery, increased oxygen demand, or a combination of the two. Decreased oxygen delivery may be encountered as the result of vascular insults (thromboembolism, vasculitis, etc.) or hypoventilation (e.g., asphyxiation). Increased oxygen demand is generally seen in hypermetabolic tissues such as neoplasms and infections.

Myo-inositol (mI): Found primarily in astrocytes, mI is a glucose-like metabolite that helps regulate cell volume. If one considers mI an astrocyte marker, then elevation of the mI ¹ H-MR spectral peak would be seen in areas of glial cell proliferation, as can be seen in Alzheimer's disease or gliosis. Conversely, lower mI spectral peaks would be observed in conditions producing astrocyte destruction (e.g., neoplasm, ischemic injury, infection).

Lipids: Lipids are found generally in the form of fatty acids, triglycerides, and phospholipids incorporated into cell membranes and myelins. They are not generally detectable on ¹ H-MRS until released by destructive processes of the brain including necrosis, inflammation, or infection. Care must be taken when selecting the volume to be studied such that extracerebral sources of lipids do not appear on the ¹ H-MR spectrum and complicate its interpretation.

Clinical Application of ¹ H-MRS

Case 1 --A 32-year-old man presented to the Emergency Department (ED) with a generalized seizure. His work-up led to the diagnosis of a left frontoparietal lobe oligodendroglioma. At the time of diagnosis, he received external beam radiation therapy and was placed on anticonvulsants. For the following 8 years, he was neurologically and radiologically stable, experiencing only occasional focal seizures involving the right lower extremity. At the time of his most recent presentation, he had been experiencing more frequent focal seizures as well as mild expressive aphasia and memory loss. MRI examination, at that time, revealed a slight increase in the size of the left frontoparietal lobe lesion with no significant change in its contrast enhancement pattern. A Thallium-201 SPECT study (figure 4A) was performed which revealed no areas of abnormal uptake. PET imaging (figure 4B) showed the left frontoparietal lesion to be hypometabolic. Both of these studies were essentially equivocal as their results could be due to postradiation scarring and/or low-grade neoplasm. Therefore, single-voxel ¹ H-MRS was obtained (figure 4C).

In this spectrum, the Cho peak is increased and the NAA peak is decreased. In addition, there is an abnormal lipid/lactate peak. These findings, together, are consistent with the presence of neoplastic tissue. The spectra obtained with a TE of 144 ms (not shown) revealed similar findings except for inversion of the lactate/lipid peak. The patient had his recurrent tumor surgically resected.

Case 2 --A 36-year-old woman had a biopsy-proven oligodendroglioma resected from the left frontal lobe. Routine follow-up MRI revealed blood products and minimal enhancement in the resection bed. Concern was raised for recurrent or residual neoplastic tissue. A multi-voxel ¹ H-MRS was then obtained (figure 5).

These spectra demonstrate the use of multi-voxel ¹ H-MRS. The multiple spectra obtained from the VOIs outside of the resection bed (numbers 16, 17, 21, 22) clearly show normal metabolite peaks; in other words, normal brain parenchyma. As the VOIs approach the lesion, there is a noticeable drop in the normal metabolite concentrations. Once in the center of the resection bed (numbers 9, 10, 14, 15), no discernible peaks are identified, a finding consistent with debris/hematoma. In all of the spectra, there are no elevated Cho peaks or decreased NAA peaks to suggest the presence of neoplastic tissue. Finally, spectra obtained with TE of 288 (not shown) did reveal small lactate/lipid peaks, indicating dead or dying tissue.

This patient was felt to have no evidence of residual or recurrent neoplasm. She received no further aggressive treatment and has continued with her routine follow-up.

Case 3 --A 32-year-old, HIV-positive man presented to the ED with a 3-month history of intermittent headaches, dizziness, and possible seizures associated with olfactory auras. In the few days prior to his presentation, he developed increasing confusion, slurred speech, and a low-grade fever. An initial computed tomography (CT) scan revealed multiple, bilateral, space-occupying lesions. Subsequent MRI showed numerous ring-enhancing lesions with associated surrounding edema. The largest of these was in the left, anterior temporal lobe. Single-voxel ¹ H-MRS was obtained through this region (figure 6).

There is substantial loss of height of the NAA peak as well as a very pronounced lactate/lipid peak. There is no significant elevation of the Cho peak. These findings, while not definitive, are strongly suggestive of an infectious process associated with tissue destruction. They do not support the diagnosis of neoplasm.

Given the patient's HIV status, toxoplasmosis was suspected; primary central nervous system lymphoma was felt to be unlikely. Although no biopsy was performed, the patient was started on anti-toxoplasma therapy with resultant improvement of his symptoms.

Case 4 --A 35-year-old woman presented to the ED with a 1-week history of bronchitis and a 2-day history of progressive neurologic decline, including aphasia and hemiplegia. An initial CT scan revealed a large, low-density, space-occupying lesion in the left frontal lobe. MRI was performed for assessment of a suspected underlying tumor. This showed a large, ring-enhancing lesion of the left frontal lobe, which crossed the midline via the superior corpus callosum. Associated edema and mass effect were also seen (figure 7A).
Single-voxel ¹ H-MRS was obtained for further evaluation (figure 7B).

Interpretation of this spectrum is not straightforward due to suboptimal water suppression. This is manifested as a rise in the spectrum's baseline as it moves towards the water peak (4.7 ppm). The ¹ H-MRS of a tumor would be expected to demonstrate an elevation of the choline peak. In this case, because of the poor water suppression, there appears to be a slight elevation of the choline peak; but this rise is, in fact, artifactual. Essentially, both the choline and creatine peaks are
normal and there is no significant decrease of the NAA peak. The most important finding is the elevated lipid/lactate peak, indicating the presence of byproducts of anaerobic respiration. This constellation of findings speaks against neoplasm and, instead, is suggestive of an inflammatory/ encephalitic etiology.

Subsequent stereotactic brain biopsy confirmed the presence of an acute inflammatory, demyelinating process; no neoplastic cells were identified. Despite appropriate treatment, the patient deteriorated rapidly and succumbed to her disease approximately 1 week following her presentation. At autopsy, the final diagnosis was acute demyelinating encephalomyelitis.

Case 5 --A 72-year-old man presented to the ED with a several month history of personality changes. During the week prior to his presentation, he had experienced progressively worsening headaches. On the morning of his presentation, he experienced new right-sided weakness. An initial CT scan (figure 8A) revealed a hypodense, space-occupying lesion in the left thalamus. MRI was requested for evaluation of a suspected tumor. It revealed abnormal T2 prolongation throughout the basal ganglia and thalami, bilaterally, extending into the deep white matter of both cerebral hemispheres (figure 8B). It also revealed thromboses of the Vein of Galen, internal cerebral veins, left transverse, and straight sinuses (confirmed by MR angiogram/MR venogram). In addition, there was evidence of new hemorrhage in the region of the left thalamus. No significant enhancing lesions were identified. Single-voxel ¹ H-MRS was performed (figure 8C).

The ¹ H-MR spectrum demonstrates considerable loss of the NAA, choline, and creatine peaks. In addition, there is an overwhelming lactate/lipid peak (notice the reversal of this peak on the TE of 144 spectrum when compared with the TE of 288 spectrum). These findings are strongly suggestive of infarction, not tumor.

The etiology of the patient's multiple thromboses was never established. However, his diffuse, acute infarctions were felt to have resulted from them. He died several days after presentation.

Conclusion

Relative ease of acquisition has made ¹ H-MRS a useful tool for the evaluation of neurologic disease. Its interpretation is based on current knowledge of the complex metabolic and biochemical processes associated with pathology. As understanding of these processes becomes clearer, interpretation of ¹ H-MR spectra will become more complex; but the spectra will also have the ability to provide more specific information. Additionally, applications of ¹ H-MRS will likely extend to involve other organ systems.

Acknowledgement

The author would like to thank Dr. Emanuel Kanal for his support and assistance with this manuscript.