The first observations that used 1
H MRS to detect differences in metabolite levels between normal brain and tumor were performed over 15 years ago (124
). Parameters that are observed in the brain at a moderate echo time of 144 ms and that provide insight into changes associated with tumor progression include levels of choline-containing compounds (Cho), creatine (Cr), N
-acetylaspartate (NAA), lactate (Lac) and lipid (Lip). The intensity of the Cho signal reflects changes in membrane synthesis and turnover that are associated with cell proliferation and remodeling. The Cr peak is often used as a reference for normalizing the intensity of other metabolites, and includes both creatine and phosphocreatine. NAA is a marker of normal brain tissue that is typically assumed to correspond to the presence of actively functioning neurons. Lac is an end-product of anaerobic metabolism and may therefore reflect ischemia and/or hypoxia. In cases in which there has been care to avoid artifacts caused by inadequate suppression of the signal from subcutaneous Lip, the presence of Lip peaks is interpreted as being a result of necrosis.
H MRS with either point-resolved spectroscopy (PRESS) or stimulated echo acquisition mode (STEAM) localization is typically applied with a voxel size of 4–8 cm3
and an acquisition time of 2–5 min (128
). MRSI provides a two- or three-dimensional array of spectra from 1–2-cm3
voxels and may use PRESS localization from a much larger selected volume or a slice selection and spin echo sequence (131
). Additional Lip suppression is typically achieved with spatially selective saturation pulses and/or an inversion pulse that nulls out these signals. At the field strength of 3 T, with a multichannel radiofrequency coil and a combination of phase encoding and echo planar sampling, it is possible to routinely obtain a 16 × 16 × 16 array of spectra from the brain with a nominal voxel size of 1 cm3
in 5–10 min (134
). Parallel imaging strategies can further cut down the acquisition time or increase coverage, but the signal-to-noise ratio is a limiting factor (137
). Although most scanners have automated shim routines, regions close to the sinuses and brainstem may still give poor quality spectra.
Numerous studies have shown that brain tumors are characterized by an increase in Cho and decrease in NAA relative to normal (139
), and that the changes in these metabolites are more sensitive to the detection of infiltration and residual disease than is gadolinium-enhanced MRI. This is particularly important for planning and assessing the response to radiation and other focal therapies (143
). An example is shown in , where the shaded voxels that have high Cho and low NAA are seen to extend far beyond CEL. To more accurately describe the metabolic lesion and make visual comparisons with the anatomic lesion, we have defined a z
score that is termed the Cho to NAA index (CNI), and have used it to quantify the difference between tumor and normal (151
). Color maps of the CNI or contours of regions with CNI greater than two or greater than three can then be superimposed on the anatomic images to represent the spatial extent of the lesion. It should be noted, however, that there are treatment effects on metabolism in normal brain (152
), and there are other disease processes, such as inflammation, that can cause a reduction in neuronal function and an increase in cellularity. This means that abnormal CNI should be viewed as defining the spatial extent of abnormal metabolism, and is consistent with, rather than specific for, the tumor.
Figure 6 Patient with a newly diagnosed glioblastoma multiforme (GBM) who had an extensive resection that left limited enhancing tumor on the post-gadolinium T1-weighted axial and coronal images (left), but for which there was a large residual metabolic lesion (more ...)
Other parameters of interest that provide useful information and can add to the specificity of CNI are the levels of Lac and Lip. With an echo time of 144 ms, the Lac peak is inverted and its presence in the tumor can be used to infer the likelihood of it being high rather than low grade. Lip peaks are at a similar frequency and may be present in grade IV glioma, even when there is no obvious area of necrosis in the anatomic images. The use of a Lac editing pulse sequence is important to avoid the signals from these peaks cancelling each other out and to unambiguously distinguish between them (155
). In vivo
MRS with shorter echo time (TE =30–40 ms) and recent studies using ex vivo
MRS have indicated that levels of myo-inositol/Cho may also be valuable for separating low-grade from high-grade glioma and in distinguishing gliosis from recurrent grade IV glioma (156
). Tissue and preclinical studies have also confirmed that phosphocholine (PC) represents the predominant contribution to the in vivo
Cho peak in grade IV glioma, whereas glycerophosphocholine (GPC) is the major component in grade II glioma (94
). Previous studies using 31
P MRSI have also demonstrated differential changes in PC and GPC in response to therapy, but the spatial resolution of 4–8 cm3
that can be obtained at clinical field strengths requires improvement.
The fact that there is abnormal metabolism in the non-enhancing region of the tumor means that it can be used to plan where tissue samples should be taken for most accurate diagnosis during biopsy or surgical resection (129
). This may be an important factor in distinguishing untreated grade III from grade II glioma, and targeting the highest CNI within the nonenhancing lesion is a sensible strategy (72
). For patients with newly diagnosed grade IV glioma, higher levels of Lac and Lip in the region with abnormal CNI have been found to be associated with worse overall survival, even when controlling for the volume of CEL (74
). This suggests that lesions that are both highly cellular and have regions of hypoxia and necrosis have a more malignant phenotype. This was found to be true for MRSI parameters obtained at both pre-surgical and pre-RT examinations (74
The spatial extent of the metabolic lesion can also be used to plan focal therapy, such as external beam RT and gamma knife radiosurgery (36
), and to assess the response to therapy. shows a patient with a grade IV glioma who had a relatively small residual CEL prior to RT, but much more extensive T2L and metabolic lesions. At the post-RT and 2-month follow-up examinations, the surgical cavity became larger and there were changes in CEL that might have been mistaken for tumor progression. In both cases, the metabolic lesion, as defined by the number of shaded voxels with CNI greater than two, decreased. At the 4-month follow-up, there was continued resolution of the metabolic lesion and reduction in CEL.
Figure 7 Changes in post-gadolinium T1-weighted images and MRSI data for a patient with a newly diagnosed glioblastoma multiforme (GBM) who was treated with external beam radiation therapy. The voxels shaded in gray have elevated choline to N-acetylaspartate (NAA) (more ...)
shows the opposite example of a patient with a newly diagnosed grade IV glioma who had a gross total resection that showed an increase in T2L at the post-RT examination, but minimal change in CEL. By the 4-month follow-up, both T2L and CEL had become dramatically larger and the patient was designated as having tumor progression. It is clear from the MRSI data obtained at these time points that there is a substantial metabolic lesion at the pre-RT examination, which becomes larger at the post-RT examination and during follow-up. It should be noted that much of the area corresponding to CEL on the 4-month follow-up examination is necrotic, with the medial nonenhancing region having the highest Cho. These two examples show quite clearly that the information provided by MRSI data is complementary to the anatomic images, and may be more valuable than CEL in assessing treatment effects.
Figure 8 Serial post-gadolinium T1-weighted images (top) and MRSI data (bottom) from a patient with a newly diagnosed glioblastoma multiforme (GBM) post-surgery and pre-radiation therapy (RT) (a), post-RT when there was no change in enhancement but some areas (more ...)
The differences in metabolite levels between gliomas with different histology can also be used to infer whether recurrent grade II gliomas have transformed to a higher grade. shows an example of a patient with an original diagnosis of grade II astrocytoma who was scanned immediately prior to surgery for suspected recurrence. The presence of diffuse enhancement on the T1 post-gadolinium images, the moderate intensity on FLAIR and ADC images, and the increased rCBV in the lesion are suggestive of a more malignant phenotype. The metabolic signature of the lesion, as shown from the Lac-edited MRSI data, includes a large region with substantially increased Cho and reduced NAA, together with a central region that also shows Lip and Lac peaks. These are all characteristics of grade IV glioma. Histological analysis of tissue samples taken from regions with elevated CNI during image-guided surgery confirmed this diagnosis.
Figure 9 Post-gadolinium T1-weighted (a), fluid-attenuated inversion recovery (FLAIR) (b), apparent diffusion coefficient (ADC) (c) and relative cerebral blood volume (rCBV) (d) images from a region of suspected recurrence in a patient who was originally diagnosed (more ...)
Another emerging MR metabolic imaging technology uses hyperpolarized 13
C agents to dramatically increase the sensitivity of the signal observed (159
). The application of low temperatures and dynamic nuclear polarization, together with a rapid dissolution process, generates a sample that can be injected into living subjects. By providing a 10 000–50 000-fold signal enhancement, it is possible to observe the delivery and flux of endogenous, nontoxic and nonradioactive substances, such as pyruvate, through key biochemical pathways, such as glycolysis, the citric acid cycle and fatty acid synthesis (162
). Preliminary studies have confirmed that 13
C-labeled pyruvate is delivered to tissues and converted to alanine, Lac and bicarbonate, with a spatial distribution and time course that varies according to the tissue of interest (163
). This is of particular interest for the management of patients with glioma because the presence of Lac is a characteristic of high-grade glioma and is a prognostic factor for poor overall survival. Preliminary results in U-87 and U-251 tumors implanted into rat brain have shown that there is significantly higher 13
C-labeled Lac in tumor than in normal brain (166
), and have demonstrated a reduction in the observed Lac within 1–2 days following treatment with temozolomide (167
). This is likely to be critical for the identification of metabolically active Lac and hence for predicting whether a low-grade lesion has upgraded, and for assessing the response to therapy for patients with glioma.