Advanced Imaging of Adult Brain Tumors with MRI and PET

Published on 09/04/2015 by admin

Filed under Hematology, Oncology and Palliative Medicine

Last modified 09/04/2015

Print this page

This article have been viewed 2571 times

4 Advanced Imaging of Adult Brain Tumors with MRI and PET

Geoffrey S. Young, Jan Stauss and Srinivasan Mukundan

Abstract

We briefly review the major advanced MRI techniques used in clinical brain tumor imaging: diffusion-weighted imaging (DWI), perfusion-weighted imaging (DSC PWI), dynamic contrast-enhanced T1 permeability imaging (DCE T1P), diffusion-tensor imaging (DTI), and MR spectroscopy and spectroscopic imaging (MRS and MRSI, respectively). These techniques provide information about tumor cellularity (DWI), white matter invasion (DTI), metabolic derangement including hypoxia and necrosis (MRS), neovascular capillary blood volume (DSC PWI), and permeability (DCE T1P). These techniques, when combined with conventional anatomical imaging (T2WI, FLAIR T2WI and delayed gadolinium [Gd] enhanced T1WI MRI) may improve diagnostic yield by providing insight into the underlying physiology described in previous chapters. Application of these techniques in clinical primary brain tumor diagnosis, grading, therapeutic planning, and monitoring of therapy are introduced, concentrating on high-grade malignant infiltrative glioma (HGG).

Introduction

As with conventional sequences, image contrast in the advanced MRI techniques is also derived from fundamental properties, such as spin density, T1, T2, T2,* and contrast enhancement of the underlying tissues. Thus, advanced techniques are also affected by the same artifacts as conventional pulse sequences; moreover, additional artifacts specific to the advanced acquisition and data processing are also seen. By subjecting the tissues to additional magnetic gradients and or radio-frequency pulses beyond what is typical for the fundamental pulse sequences, the signal emitted by the tissue is modulated, resulting in images with advanced tissue contrasts that are more reflective of relevant functional characteristics such as blood volume, permeability, hypercellularity, etc. Unfortunately, the price paid for this additional information is reduced signal-to-noise ratio and spatial resolution in comparison to the conventional anatomical images. For these reasons, it is critical that the neuroimager interpret information from advanced MRI sequences in the context of high-resolution conventional MR examination. Frequently, in fact, advanced sequences are co-registered to high-resolution conventional Gd enhanced T1WI with 1 mm or smaller voxel size to facilitate better interpretation.

Knowledge of the conventional morphological MRI and CT appearance of brain tumor forms an essential foundation for the interpretation of advanced imaging. Although it is beyond the scope of this chapter, it seems reasonable to bring to the reader’s attention one recent morphologic imaging report that, pending prospective validation, suggests that in anaplastic astrocytoma an expansile growth pattern and lack of enhancement may predict longer survival.¹ In addition, this clinical introduction will not review the more than half-century-long literature on MR physics, engineering, and basic tissue physiology that has preceded the recent clinical application of these techniques. This background understanding is not essential for the neurologist, oncologist, or surgeon caring for brain tumor patients, but is critical for the neuroimaging researcher and practitioner. Numerous other advanced MRI techniques are excluded from this review because they have not yet achieved widespread clinical application. Finally, the reader is cautioned that the rapid progress of commercial MRI hardware, software, and clinical literature necessitates frequent reevaluation of the conclusions of this review.

Increasingly widespread research use of advanced brain imaging techniques in humans and animal models is making a great contribution to scientific understanding of brain tumor pathophysiology, to evaluation of new therapies, and in discovery of predictive markers that promise to assist in HGG phenotyping for personalization of therapy. These vital and exciting efforts, should not be confused with the clinical goals of advanced brain tumor imaging, which primarily focus on four issues that present significant challenges for conventional MRI:² (1) differentiation of primary infiltrative glioma from other primary brain tumors, metastatic tumors, strokes, infection, and tumefactive demyelination; (2) preoperative grading of primary glioma; (3) planning of biopsy, resection, and radiation therapy, including detection of tumor margin; (4) sensitive early detection of recurrence or progression in HGG, with distinction from radiation necrosis. In particular, in the last few years, the problem of detecting recurrence has become far more difficult because of the widespread use of temozolomide. This has increased both the incidence of postradiation enhancement mimicking progression (pseudoprogression) in approximately 20% of treated patients³ and that of angiogenesis inhibitors. The latter appear to alter the pattern of recurrence and progression by decreasing enhancement without definitely prolonging survival.^4,⁵ Advanced MRI aims to address these challenges by characterizing and monitoring the four most important independent pathophysiologic attributes of each individual patient’s brain tumor: (1) hypercellularity, (2) high invasiveness, (3) hypermetabolism, (4) hypervascularity. These four attributes are known as the “4-Hs.” The pathophysiologic heterogeneity of brain tumors makes assessment of all 4-Hs crucial to a modern multiparametric brain tumor characterization. All HGGs have a least one of these attributes, and a few will have all four. Thus assessment of all 4-Hs is essential, at least at baseline.

Diffusion Weighted Imaging (DWI) of Cellularity

DWI image contrast is based on random thermal diffusion (Brownian motion) of water molecules in each voxel of brain tissue. In bulk water, the movement (diffusion) of a given water molecule is not constrained by boundaries, and the average diffusivity (diffusion per unit time) is proportional to the temperature. At higher temperatures, molecules have greater Brownian motion and therefore have greater diffusivity. In brain, the temperature is constant, and the distance water diffuses during a fixed time is mainly determined by physical constraints to water diffusion at the cellular and subcellular levels. Formally, diffusivity is determined by the fraction of tissue water in the intracellular compartment, where the tightly packed membranes of intracellular organelles hinders the free diffusion of water. In comparison, water in the extravascular extracellular space (EES) has a diffusivity (ADCe) that is an order of magnitude greater than the intracellular diffusivity (ADCi). Since the size of neurons is much smaller than the size of the DWI voxels, the average apparent diffusion coefficient (ADC) of each voxel is primarily influenced by the ratio of extracellular to intracellular water, referred to as the extracellular volume fraction (EVF) or local tissue “cellularity.”^6,⁷ While this oversimplification of a very complex and technically heterogeneous literature may provoke controversy among imaging experts, it is generally accepted as a first approximation. See Figure 4-1.

Figure 4-1 60-year-old man with treated gliosarcoma in the left temporal lobe. Hyperintensity in the left temporal lobe on the DWI (middle), confirmed by the ADC map (right) representing low diffusivity, suggests high cellular tumor, more marked than the degree of abnormal contrast enhancement (left).

CELLULARITY IN DIFFERENTIAL DIAGNOSIS

DWI can be crucial in helping to distinguish brain tumor from tumefactive nonneoplastic disorders. The use of DWI in ischemia is well known. This can be of significance in rare cases where mass-like presentations of embolic and vasculitic ischemia are difficult to distinguish from brain tumor.⁸ In addition, sensitivity and specificity of over 90% are routinely achieved using DWI to distinguish the low ADC of epidermoid (due to the presence of sloughed epithelial cells, cholesterol, and keratin) from the high ADC of pure CSF-containing arachnoid cyst. Similar accuracy is achieved in distinguishing the low ADC of abscess, filled with white blood cells, from the high diffusivity of CSF-filled necrotic tumor cavities. A rim of low diffusivity at the periphery of a lesion may also be helpful in suggesting tumefactive demyelination.⁹^–¹²

DWI can also be useful in preoperative differential diagnosis of brain tumors. In intra-axial tumors, low ADC suggests that lymphoma, medulloblastoma, or metastasis should be considered; the high cellularity of these lesions typically results in a much lower ADC than for HGG.^13,¹⁴ Similarly, low diffusivity in extra-axial masses suggests highly cellular meningioma or dural metastasis. Nevertheless, glioblastoma and gliosarcoma cannot be completely excluded; a small number are very cellular and present with a high ADC that overlaps with that of the other tumor types.^13,¹⁵^–¹⁷ This illustrates the most important biological insight and, at the same time, the most important caution critical to responsible clinical use of advanced imaging: because HGG genetics, pathophysiology, and imaging phenotypes are so diverse, no single image data subtype can reliably be interpreted in isolation from the other advanced imaging data, conventional imaging data, and clinical history.

CELLULARITY IN TUMOR GRADING AND THERAPEUTIC PLANNING

A number of studies support the correlation of low minimum ADC (ADC_min) with high cellularity in tumors, including low-grade glioma, high-grade glioma, medulloblastoma, lymphoma, meningioma, and metastasis.^13,¹⁸^–²² Lower ADC correlates with atypical and malignant pathologic subtypes of meningioma, but the ADC overlap between low and high grade populations is too great to allow reliable prediction of tumor pathology or behavior in individual patients.²¹ In part, this seems likely to reflect the importance of vasogenic edema produced by vascular endothelial growth factor (VEGF) secretion and tissue invasion in determining meningioma behavior. Several groups have reported that ADC_min less than 1.7 to 2.5 × 10^–3 mm²/s within the cellular portion of glioma correlates with high grade.^23,²⁴ ADC varies significantly within each grade, especially among HGG.^14,^23,²⁵^–²⁷ In addition to variation in cellularity, variation in the degree of necrosis, hemorrhage, and calcification likely contribute to this finding, as does variation in vascular permeability, related to angiogenesis and secretion of VEGF and other vasoactive paracrine factors. Although this variation reduces the likelihood that DWI alone can reliably predict histopathology, it suggests that ADC or other metrics derived from DWI may help in prediction of response to radiation.²⁸

DWI MONITORING OF THERAPEUTIC RESPONSE

Detection of low ADC at the surgical resection margin on immediate postoperative imaging should suggest marginal ischemic necrosis rather than residual tumor.^29,³⁰ While persistently low or decreasing ADC within the cavity or extra-axial space should suggest the possibility of pyogenic infection, the temporal evolution of postoperative hematoma and necrotic debris often produces a complex DWI and ADC appearance, so careful correlation with changes over time in the imaging studies and clinical presentation is essential.³¹

Beyond detection of these postoperative complications, low ADC offers an independent parameter for predicting malignancy and aggressive behavior in gliomas. In patients whose preoperative MRI demonstrates an atypical MRI pattern, low ADC evidence of high cellularity predicts aggressive clinical behavior and may, in some cases, be a better predictor than histopathology.³² Although published thresholds vary, it has been shown that patients with minimum intratumoral ADC (minADC) less than 1.0 × 10^–3 mm²/s have a much worse prognosis than those with higher-ADC tumors.³³ Because EPI DWI-derived ADC estimates vary greatly with instrument, precise acquisition parameters, and postprocessing, investigations of normalized ADC ratios would seem to be indicated and, indeed, are beginning to be published.

In tumors with high baseline cellularity, such as highly cellular GBM or medulloblastoma, ADC may aid in early detection of treatment response to chemoradiation, as cellularity decreases in response to cytotoxic chemoradiation.^19,^20,³⁴^–³⁷ Because of interscan variation in ADC estimates, normalized ADC ratios (nADC) may prove more robust than absolute ADC measurements in separating radiation necrosis and pseudoprogression from recurrence after XRT. These measures are attracting increasing interest, since ADC in brain tumor seems relatively less affected by steroid use and angiogenesis inhibition than are enhancement, tumor edema, and permeability.³⁸ As suppression of enhancement by angiogenesis inhibition makes follow-up with conventional enhanced imaging less reliable, ADC may offer an important complement to blood volume imaging for longitudinal follow-up. This promise justifies ongoing development of higher b-value and multiple b-value echoplanar (EPI) sequences as well as non-EPI techniques to improve estimates of ADC, estimates of cell volume fraction. and longitudinal registration. “Functional diffusion mapping” is one investigational method of quantifying longitudinal change in ADC, among a number of such methods under study.^39,⁴⁰ See Figure 4-2.

Figure 4-2 In a patient with GBM, the presurgery MRI (top row) shows very low diffusivity on ADC map (middle) and DWI (right), corresponding to abnormal gadolinium enhancement (left). While high doses of perioperative steroids decrease the permeability and eliminate contrast enhancement almost completely, they do not affect the cellularity, which is reflected by persistent reduced diffusivity as is seen on the postsurgical MRI (bottom row). This illustrates the utility of DWI in follow-up of patients on steroids and angiogenesis inhibitors, both of which strongly suppress permeability and enhancement.

White Matter Invasion Assessment with Diffusion Tensor Imaging (DTI)

DWI, by imaging the motion of water in one to three spatial directions, acquires enough information to estimate the magnitude of random thermal diffusion. DTI acquires information in at least six directions, and completely defines a tensor (three-dimensional vector) describing both the magnitude and direction of water diffusion.^41,⁴² A large number of techniques that acquire many more diffusion directions to much more precisely define local water diffusion direction have been published under various names (q-ball imaging, diffusion spectrum imaging, etc.). Precise detection of the preferred direction of water diffusion is of interest in brain imaging because the myelinated axon bundles assist water diffusion along white matter tracts and prevent water diffusion across the tracts, a tissue property referred to as “anisotropy.” The relative degree of white matter “diffusion anisotropy” in each voxel can be characterized by a number of derived scalar metrics comparing diffusivity in one direction with another, of which the most widely used is termed the fractional anisotropy (FA).

DTI TRACTOGRAPHY FOR SURGICAL GUIDANCE

Alternatively, the exact vector diffusion tensor can be depicted graphically in each voxel and the voxels connected to depict the mean fiber orientation in various major white matter tracts. A large number of algorithms for producing such “tractograms” have been published.

In conjunction with blood oxygen level dependent (BOLD) functional MRI (fMRI) depiction of sensorimotor, visual, and primary language cortical activation, DTI tractography have evolved into robust techniques for detecting the location of critical white matter tracts that are next to, displaced by, or invaded by tumor, allowing avoidance of operative injury to functional tracts and prediction of postoperative disability that may result from their transection.⁴³^–⁴⁷

DTI Assessment of White Matter Invasion

Interest in imaging the white matter infiltrative component of HGG continues to increase because temozolomide chemoradiation and angiogenesis inhibition have been shown to alter the pattern of recurrence ⁴ by improving control of the enhancing solid component. FLAIR T2WI is very sensitive to the presence of vasogenic edema elicited by microscopic tumor invasion, but does not allow reliable distinction of direct tumor infiltration from peritumoral edema. A number of studies suggest that a decrease in FA or other DTI-derived measures of white matter anisotropy may be a marker for white matter disruption due to local glioma infiltration, as other causes of vasogenic edema would not be expected to actually disrupt white matter tracts. Observation of lower white matter anisotropy near to the high-grade tumor masses was encouraging; this remains an exciting research focus,⁴³ although one not yet ready for translation to clinical use. Reports of widely different results from groups employing different combinations of angular resolution, b-value, and signal-to-noise ratio (SNR) illustrate that acquisition and postprocessing techniques will need to mature before clinical trials demonstrating effective detection of the margins of WM invasion are likely to be successful.^27,⁴⁸^–⁵⁴ Unfortunately, since no effective treatment for infiltrative tumor exists, biopsy of involved white matter is difficult to justify ethically. Many resections are performed with suction, which makes it difficult to track the origin of the tissue; also, no widespread robust method has been developed to correlate tissue with MRI on a millimeter scale. Therefore, human DTI translation research remains hampered by lack of a valid gold standard. A successful alternative strategy has been to use DTI to predict clinical outcome; classification of preoperative glioma margins as infiltrative or expansile by qualitative interpretation of DTI tractography correlates with survival.⁵⁵ See Figure 4-3.

Figure 4-3 Post-Gd T1WI (top left) depicts a focal ring-enhancing glioma with peripheral high blood volume on the CBV map (top right). Although hypervascular, this glioma is expansile rather than infiltrative, as is seen on the tricolor FA map (top center). The infiltrative-appearing glioma on the FLAIR image (bottom left), in contrast, has low blood volume (bottom right), but while hypervascular, has an infiltrative pattern on the FA map (bottom center).

(Figure courtesy Shuohui Yang, MD and X. Joe Zhou, PhD.)

Another strategy may be to try to demonstrate the efficacy of DTI-derived anisotropy, in combination with ADC, as an early indicator of response or potential survival in patients undergoing chemoradiation.⁵⁶ Such studies are ongoing using a large number of more sophisticated metrics of white matter coherence. These measures promise to achieve greater sensitivity and specificity by exploiting the tensor directional information from DTI and DSI more fully than simple FA.^48,^51,⁵⁷^–⁵⁹

Spectroscopy of Tumor Metabolic Derangement

MR spectroscopy (MRS) and spectroscopic imaging (MRSI) essentially represent in vivo application of nuclear magnetic resonance (NMR). NMR revolutionized analytic chemistry in the 1940s by allowing chemists to nondestructively assay the chemical composition and bond structure of organic molecules. Differences in the spin density of the surrounding electron cloud produce a different degree of magnetic shielding at each chemically unique position in a molecule. This “chemical shift” alters the applied external magnetic field experienced by each proton in a molecule, causing it to precess at a slightly different frequency and emit a slightly different frequency of radio waves when excited. The radiofrequency is detected and plotted on a graph in which the x-axis displays the spectrum of frequencies emitted by the sample in parts per million (ppm) relative to a standard reference, and the y-axis displays the magnitude of each frequency in arbitrary units relative. The use of ppm rather than hertz (Hz) allows spectra acquired at different field strengths to be directly compared.^60,⁶¹ Different chemical compounds are identified in the spectrum (“assigned”) by recognition of one or more peaks representing the resonance from distinct species of protons in that compound. The spectrum can be acquired from a single voxel (MRS), a two dimensional matrix of voxels from a single slice (2-D MRSI), or a rectangular three-dimensional matrix of voxels (3-D MRSI). MRSI data are sometimes displayed as color “metabolite maps” that give a qualitative impression of the anatomic distribution of the height, area, or ratio of height or area corresponding to important peaks or assigned compounds, but interpretation relies principally on inspection of the graphed spectra.

Although MRI and NMR are in principle the same, certain physical and practical limitations of clinical human MRS make it unlikely that human MRS will ever completely fulfill the high hopes that surrounded initial implementation over 20 years ago. These limitations include: much lower achievable field strength (3T vs. 20+T), much higher sample temperature (37 ° C vs. freezing in liquid nitrogen), much shorter tolerable scan times (less than 10 minutes vs. hours or days), rarity of MR detectable spin 1/2 nuclei other than protons (¹H₀) in human tissue, inability to perform heavy isotope labeling in vivo or centrifugally spin patients, and extremely complicated mixtures of metabolites in tissue. Nevertheless, progress has made MRS valuable for a number of important niche applications in clinical neuroimaging care and research.

The most important assigned peaks observed with 1.5 T to 3.0 T in vivo MRS of brain tumor patients are: branch chain amino acids produced by lysosomal catabolism in activated polymorphonuclear leukocytes (PMN) (AA: 0.9–1.0 ppm), lipid products of necrosis (Lip: 0.9–1.5 ppm), lactate from anaerobic glycolysis (Lac: 1.3 ppm), alanine (Ala: 1.5 ppm), n-acetyl aspartate associated with intact neuronal membranes (NAA: 2.0 ppm), choline released during cell membrane synthesis or degradation (Cho: 3.2 ppm), energy storage creatine compounds (Cr: 3.0 ppm and 3.9 ppm) and myoinositol (mI: 3.6 ppm). Lipid and lactate peaks represent a number of compounds with similar structures and so produce broad peaks, and creatine produces two easily detectable peaks corresponding to two chemically nonequivalent species of protons. AA and Lac can be differentiated from the overlapping broad Lip peak, when needed, by acquiring spectra at different TEs, since the protons forming these two peaks precess out of phase with the Lip, NAA, Cr, and Cho peaks. Analysis of these major assigned peaks in brain spectra can provide important information about pathophysiology but not etiology. Decreased NAA is seen with neuronal injury of any cause, and increased Cho with glial growth or injury of any cause; Lac appears with all causes of anaerobic glycolysis, and increased Lip and decreased Cr with all causes of necrosis.^62,^63,⁶⁴

Spectroscopy in Differential Diagnosis

Elevated Cho and decreased NAA with variable occurrence of Lac and Lip represent the typical spectra seen in glioma, but identical abnormal spectra may be seen in ischemia, demyelination, infection, and other pathologies. For this reason, MRS is not generally useful in differential diagnosis of brain masses; however, it has been carefully studied in a number of niche applications, such as distinction of meningioma from dural metastasis and peripheral HGG. All three may have very high Cho, but because meningioma and metastases contain no neurons, the spectra demonstrate no detectable NAA. However, this signature is not definitive, since focal HGG may also contain no detectable NAA. The addition of Lac and Lip in this context favors HGG, but may also occur in metastasis. Similarly, the small ALA peak detected in 80% of meningioma spectra is found in roughly the same proportion of metastases and schwannomas.⁶⁵ Nevertheless, observation of very high ALA in the context of characteristic PWI, DWI, and basic imaging patterns may help to distinguish meningioma, peripheral GBM, gliosarcoma or other intra-axial tumors. Similarly, extensive clinical research has failed to support early hopes that MRS would help to distinguish tumor from tumefactive ischemia and demyelination.⁶⁶^–⁶⁹ As noted, the reason for this failure is that the principle peaks in MRS reflect pathophysiology rather than etiology: rapid breakdown of glial membranes in ischemia and demyelination releases as much choline as rapid membrane synthesis in HGG, neuronal injury reduces the NAA peak regardless of the cause, and Lac and Lip derived from anaerobic glycolysis and necrosis are common to many pathologies. Nevertheless, in adult neuroimaging there is one clinically useful application of MRS in differential diagnosis: distinction of bacterial, fungal, or parasitic abscesses from cystic necrosis due to tumor or radiation by detection of the AA peak specific for presence of activated PMNs.^10,⁷⁰ See Figure 4-4.

Figure 4-4 One voxel from a 2-D MRSI (top center) of infiltrative glioma seen on FLAIR image (top left) demonstrates an increased Cho/NAA ratio, suggesting increased membrane turnover and neuronal injury typical of low-grade glioma. The voxel displayed, selected based on the MRSI color maps of NAA, CR, Cho (bottom row), contains the highest Cho/NAA ratio and was selected as the target for biopsy (top right) in order to reduce the risk of undergrading.

Spectroscopy in Glioma Grading and Biopsy Guidance

In clinical practice, MRS can help to suggest the presence of high-grade tumor in areas where Cho/NAA peak height ratios (CNR) are greater than 1.5.^24,^71,⁷²^–⁷⁴ A number of derived semi-quantitative metrics, such as CNR R-values normalized to the contralateral white matter, can be helpful in accounting for anatomic and individual variation of these ratios. One important caution is that higher NAA/Cho ratios are more often found in grade III than in grade IV glioma. For this reason, detection of Lip and Lac evidence of necrosis and anaerobic metabolism in spectra of untreated glioma can add to suspicion of WHO grade IV tumor.^24,⁷⁵ MRS is a particularly useful adjunct to PWI for preoperative grading of oligodendroglioma, since high blood volumes on PWI may be misleading.⁷⁶^–⁷⁹ Interpretation of MRS in the context of PWI and basic imaging characteristics is critical to avoiding this pitfall.

Targeting metabolically active tissue with high Cho/NAA ratios can decrease rates of “undergrading” and false negatives related to sampling error in biopsy of heterogeneous HGG.^80,⁸¹ These successes have motivated application of MRS to guide radiosurgery,⁸²^–⁸⁴ and suggest that wider utility may emerge as the automation, speed, resolution, and reproducibility of MRS gradually increases. Unfortunately, confirmation of the utility of MRS for guidance is hampered by the pathologic heterogeneity of HGG, the poor efficacy of current ablative therapies and the intractable problem of obtaining precise correlation with tissue samples.

More recent analyses have shown that whole brain ratios of NAA (WBNAA) may be decreased by up to 30%—far more than can be explained by focally detectable tumor.⁸⁵ This approach seems particularly timely since whole brain markers of infiltrative tumor burden are becoming a more significant contributor to patient mortality as control of focal recurrence improves. MRS assessment of CH₂/CH₃ ratios within the lipid spectrum of normal-appearing white matter may offer a complementary nonlocalized tumor burden assay.⁸⁶

MRS in Posttreatment Monitoring

Differentiation of recurrent or progressive HGG from early radiation effect and chemoradiation necrosis remains difficult, especially because a mixture of both processes is present in many patients. Here again MRS may prove a valuable adjunct to PWI, although its efficacy remains unproven.⁸⁷ Since Lac and Lip are seen in both processes, observation of these peaks is not sufficient to exclude recurrence unless accompanied by absence of Cho and NAA peaks, or evidence of decreasing Cho over the course of MRSI follow-up. This is particularly convincing if corroborated by increasing ADC and decreasing CBV. Similarly, increasing Cho/NAA ratios over time are a sensitive sign for early tumor recurrence.^81,^83,⁸⁸^–⁹⁰ Interpreted in the context of other basic and advanced imaging data, technically meticulous serial MRSI has been shown to be valuable in early detection of tumor recurrence. Unfortunately, because voxel to voxel variation in Cho, NAA, Lac, and Lip at each time point is usually greater than the change in spectra over time, minute differences in technique can render longitudinal comparison invalid. Similarly, very significant errors in interpretation can result from slight errors in MRS or MRSI voxel positioning resulting in accidental inclusion of small amounts of fat from skull marrow; scalp; choroid plexus; dural ossification; magnetic susceptibility artifact from bone or metal; or CSF in ventricles.

For this reason, the groups that have been most successful in use of serial MRSI have found it essential to have a neuroimager expert in MRS supervise each acquisition, a requirement that, in combination with the long time required for MRS—typically 20-plus minutes per patient—and lack of insurance reimbursement, has made high-quality, reliable MRS unachievable in routine clinical care at many of the largest centers. Hopefully, progress in automation of MRSI acquisition may address these issues in future, but for the moment, MRSI remains the advanced imaging modality most subject to individual variation in resources and expertise.