P7S Medical Review: Spring 1997, Vol.4, No.1
Advanced Neuroimaging Techniques To Distinguish Brain Tumor Recurrence From Post-Radiation Necrosis

P&S Class of 1996

Address for Correspondence:

Joel K. Erickson, M.D., Department of Radiology

Massachusetts General Hospital, Boston, MA 02114

INTRODUCTION

Neuroimaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI) are important in the diagnostic and therapeutic evaluation of brain tumors. An enhancing mass on either of these modalities suggests malignancy in the central nervous system1. If biopsy confirms the diagnosis, radiation therapy becomes one of the most important treatment modalities, either as primary therapy or adjuvant therapy following surgical resection. The goal of radiation therapy is to use targeted beams to halt progression of tumor while sparing normal surrounding tissue2. Most tumors respond to external irradiation, resulting in cures, prolonged survival, or symptom alleviation. Although irradiation of brain tumors can halt tumor growth, injury of adjacent normal tissue is often an inevitable complication. The radiation dose is determined by the risk of irradiation-induced normal brain necrosis, balanced against the risk from tumor progression. In adults with malignant tumors such as glioblastoma multiforme where survival is usually measured in months, the risk of rapid tumor recurrence is so high that a greater risk of necrosis is accepted and maximal radiation doses are routinely used. For adults with low-grade gliomas, such as astrocytomas, an incidence of necrosis greater than 5% is usually not accepted.

The imaging modalities of CT and MRI again become important following the completion of radiation therapy. A 6 month follow-up CT or MRI scan is typically obtained, followed by frequent serial imaging studies to monitor the tumor response. The sensitivity of CT and MRI in detecting a lesion is based upon breakdown of the blood-brain barrier, a pathological consequence of both necrosis and recurrent glioma. Radiation injury causes pathological changes that impair vascular integrity, whereas tumor associated angiogenesis produces structurally weak vessels with leaky membranes3. Both types of vessels allow local extravasation of infused contrast material, which appears as enhancement on radiologic studies. As a result, an area of new enhancement on post-treatment CT could represent successful tumor destruction with necrosis of surrounding tissue, tumor recurrence in the original site, or a combination of both. Distinguishing between these two is clinically important since inappropriate treatment of benign radiation necrosis might involve unnecessary additional surgery and/or irradiation, yet attributing recurrent tumor to radiation necrosis could delay or preclude needed therapy.

Since standard neuroimaging methods cannot reliably distinguish radiation necrosis from tumor recurrence, other advanced imaging modalities have been investigated which utilize different technologies. These modalities primarily involve in vivo metabolic information to permit both quantitative and qualitative estimation of metabolic activity of normal brain and tumors. Metabolic information can include glycolytic activity of cells, transmembrane transport, and blood flow in regions of interest, usually measured following the injection of radioactive analogues. The current standard modality for making the distinction between radiation necrosis and brain tumor recurrence is positron emission tomography (PET), which identifies the distribution of radiolabeled glucose in brain tissue4. Although PET has high sensitivity and specificity for radiation necrosis, it is not a foolproof method as several cases have documented misdiagnoses. Thus, other types of imaging have been studied for use as single or combined modalities with PET, including magnetic resonance spectroscopy (MRS), which identifies biochemical metabolites, and functional magnetic resonance imaging (fMRI), which provides metabolic information by detecting changes in regional cerebral blood volume. These technologies are among the most advanced in diagnostic radiology, and may ultimately prove to be highly beneficial modalities in the post-irradiation care of patients with brain tumors.

PATHOLOGIC EFFECTS OF CNS RADIATION INJURY

Understanding the pathogenesis of CNS radiation injury is helpful in interpreting metabolic neuroimaging data, which reflects the pathologic changes of both irradiation and tumor growth. The effects of CNS radiation are generally classified into three categories: acute, early delayed (0-3 months), and late delayed (9 months to 2 years)3. Acute effects of radiation include increased intracranial pressure and exacerbation of symptoms associated with tumor, which are usually caused by edema and respond to simple corticosteroid therapy. Early delayed reactions include somnolence and lethargy, usually occurring between 2 and 6 weeks following completion of therapy. The pathogenesis of the early delayed reaction is thought to be associated with a transient interruption of myelin synthesis produced by radiation injury to the oligodendroglial cells. The turnover time of myelin, 5 weeks to 2 months, corresponds to the latency and recovery time of the syndrome. Late delayed radiation reactions occur several months to years later. They are usually irreversible and progressive, and symptoms depend on the location of the lesion in the brain. Clinical manifestations of late radiation injury may include narrowing of large vessels, pituitary-hypothalamic dysfunction, and secondary neoplasia.

Figure 1. A, In the acute stage of late delayed radiation necrosis, fibrinoid necrosis occurs in which leaky small blood vessels allow for the escape of fibrin into the perivascular regions.	B, In the chronic stage, marked thickening of an arteriole is present. (Reprinted with permission from Burger PC, Boyko OB, 1991.)

The most frequent and most severe complication of late radiation injury is radiation necrosis, which tends to occur a year or more after irradiation3. The most characteristic pathological abnormality of radiation necrosis is the exudation of eosinophilic fibrin, which forms hypocellular layers along the gray-white junction. Although many small vessels in the white matter demonstrate classic fibrinoid necrosis, others allow fibrin to escape into the surrounding brain. In the chronic stage, areas of parenchymal necrosis are resorbed by macrophages, coalescing to large cysts with parenchymal dystrophic calcification. The gray-white junction fibrin exudate can disappear, leaving a thinned cortical mantle with a ribbon of underlying encephalomalacia, which may also calcify. With time, the vascular changes evolve, and the acute fibrin exudation and fibrinoid necrosis are gradually replaced by telangiectasia and vascular thickening (Figure 1).

The pathology of normal tissue radiation necrosis is distinct from the pathologic changes following irradiation of malignancy3. When a tumor is irradiated, it is often stripped of the smaller and more anaplastic elements and converted to paucicellular lesions notable for considerable nuclear pleomorphism. The residual pleomorphic tumor cells have prominent cytoplasm and conspicuous cell processes. In addition to cellular changes, radiation-induced vascular changes are also observed, including thickened vessels, fibrinoid necrosis, and abnormalities of the deep cortical laminae. Tumor recurrence, when present, is usually located most prominently along the rim of the original neoplasm within several centimeters of the original contrast enhancement on CT or MRI, and is usually composed of markedly cellular foci of small anaplastic elements. Such cells reconstitute the neoplasm to its original hypercellular state and may sometimes create a recurrent neoplasm that is more anaplastic than the original.

New areas of enhancement on standard imaging techniques following radiation therapy are frequently biopsied, yet interpretation of this tissue is difficult. At least four patterns may be seen within the same specimen of an irradiated glioma at the time of recurrence: pleomorphic and presumably "inactive" neoplasm, radiation necrosis, parenchymal gliosis, and markedly cellular recurrent tumor5. Histopathologic identification of a highly malignant neoplasm thus cannot be interpreted as indicating that the entire mass is composed of similar tissue. Likewise, the presence of radiation necrosis does not ensure that active recurrent neoplasm is not present elsewhere in the lesion. Therefore, metabolic neuroimaging offers the best insight into the characteristics and prognosis of post-radiation morphologic changes within an entire tumor bed and can help predict the overall malignant potential of the lesion6.

ADVANCED NEUROIMAGING METHODS

Histologic findings of white matter coagulation and fibrinoid necrosis are associated with radiation necrosis, whereas the proliferation of structurally weak vessels is associated with tumor angiogenesis. Each of these factors may lead to breakdown of the blood-brain barrier. Acute changes appear as non-specific contrast enhancement on CT and MR imaging, and these modalities cannot distinguish between radiation necrosis and tumor recurrence. Thus, positron emission tomography (PET), magnetic resonance spectroscopy (MRS), and functional metabolic imaging (fMRI), imaging techniques which reflect metabolic utilization of injected materials rather than contrast effects alone, have been extensively studied as non-invasive modalities for post-radiation follow-up.

Positron Emission Tomography (PET)

Positron emission tomography (PET) using the radiolabeled glucose [18F]-2-fluoro-2-deoxyglucose (FDG) may often provide the only clues for making the distinction between radiation necrosis and tumor recurrence4. PET provides data about the brain such as blood flow and blood volume, as well as metabolic information such as glucose utilization and protein synthesis. A commonly used protocol in PET involves the following: 1) the injection of radiolabeled glucose which emits positrons for detection, 2) a 40-minute waiting period to allow the agent to accumulate in tissues, and 3) acquisition of cross-sectional images which are color-coded to permit both quantitative and qualitative estimation of metabolic activity. As malignant tumors in general are associated with a high glycolytic rate, the uptake of labeled deoxyglucose into the cell may reflect their disease activity7,8. Within necrotic tissue, a very slow rate of dephosphorylation is present, rendering a smaller regional uptake of labeled deoxyglucose as compared to normal tissue9 (Figure 2).

Figure 2. PET imaging of late delayed radiation necrosis. A, Bilateral subcortical white matter enhancement is seen in a patient radiated for an astrocytoma. B, FDG-PET imaging showed only the right hemispheric lesion to be hypermetabolic (large arrow). Recurrent astrocytoma was present in the biopsy specimen. The left hemispheric lesion was hypometabolic (small arrow). Radiation necrosis was present in the biopsy specimen. (Reprinted with permission from Davis et al, 1993)

The usefulness of this technique was first established in 1987 when it was shown that a ratio of lesion FDG uptake to white matter FDG uptake greater than one was more likely to be associated with recurrent tumor, whereas a ratio of less than one was more likely to be associated with necrosis4. In 1988, a study of patients treated with interstitial brachytherapy, a treatment associated with a higher incidence of radiation necrosis, showed that active tumors have activity equal to or greater than immediately adjacent normal cortex, whereas lower lesion activity was interpreted as radiation injury10. To determine the prognostic value of the PET diagnosis, the PET result was compared to the clinical diagnosis of tumor recurrence or no tumor recurrence over the next 12 months, established by the patient's progress and response to therapy. The PET result agreed with the follow-up clinical diagnosis in 15 of 17 cases of tumor recurrence and 17 of 21 cases of radiation injury. The overall accuracy of the PET examination was 84%, and the positive and negative predictive values were 79% and 89%, respectively.

Among 18 patients from this group, a histologic diagnosis was obtained at reoperation after PET studies. It was found that tissue from the site of irradiation contained apparently viable tumor cells as well as necrotic tissue in all cases, regardless of clinical outcome. Clinically, it was clear that tumor was proliferating in some and not proliferating in others, but the two groups could not be distinguished histologically. This result suggests that although tumor cells at the radiation site may appear to be morphologically intact, they may be metabolically and clinically inactive, and a reliable biopsy diagnosis may not be possible. The functional and global nature of PET allowed visualization of whole-lesion metabolic activity, which more closely correlates with clinical tumor activity and better predicts patient outcome.

Despite the success reported for PET, there are some drawbacks to its utilization. For example, false negative scan results have been demonstrated in patients following brain tumor therapy. A 1995 study evaluated 39 patients with FDG-PET following surgery and radiation for primary brain tumors11. In this group, two patients were identified that had PET scans consistent with radiation necrosis, yet had subsequent clinical deterioration and pathologically confirmed high grade tumor. Although FDG-PET has been successful in distinguishing high-grade tumors from radiation necrosis, this distinction is not well shown in patients with low-grade tumors, nor extremely high-grade tumors associated with necrotic centers. Areas of necrosis are metabolically inactive and do not demonstrate glucose uptake; therefore, the most malignant tumors may in fact show decreased glucose uptake compared to normal tissue. Finally, PET imaging is extremely expensive and availability is limited to a small number of research institutions12. Many patients are unable for financial reasons to undergo this examination. Less costly and more accessible modalities are clearly desirable to provide affordable care for the greatest number of patients.

Magnetic Resonance Spectroscopy (MRS)

Magnetic resonance spectroscopy utilizes the same physical properties as MR imaging, and can be obtained during the same examination. MR spectroscopy acquires resonance signals from tissue nuclei with an odd number of protons and neutrons such as 1H, 13C, 23N, 31P, which have a magnetic moment and interact with the magnetic field. The information obtained from this technique is presented as a biochemical spectrum rather than an image, providing a variety of quantitative data depending on the nucleus involved. For instance, measurement of tissue pH is accomplished by 31P MRS from the resonance position (frequency) of the inorganic phosphate. The majority of neuroradiologic studies utilize 1H MRS to obtain information about the identity of metabolites (from their peak position in the spectrum) and quantification of each metabolite (from the area under the peak)13. The metabolites usually present in a standard 1H brain analysis include N-acetylaspartate (NAA), considered to be a reliable neuronal marker; choline, a constituent of cellular membrane which servers as a marker of cellularity; creatine-phosphocreatine, an energy-providing compound in cellular metabolism; and lactate, the product of anaerobic metabolism seen in areas of hypoxia14,15 (Figure 3).

In a 1992 study proton MRS was utilized to evaluate 50 brain tumor patients in order to define the spectroscopic characteristics of both malignant tissue and radiation necrosis16. The results indicated a decrease in NAA in both tumor and radiation necrosis consistent with the loss of functioning neurons in both pathologies, an increase in choline in tumor recurrence reflecting the increased cellularity of malignant tissues compared to the decreased choline associated with necrosis, and higher glucose utilization rates for lesions in which lactate was detected, usually high-grade gliomas. The various biochemical components provided strong evidence for the presence of tumor since increased choline and a lactate peak would not be expected in necrotic tissue. The authors concluded that choline is the most important metabolite for the distinction between tumor and radiation damage, particularly if a pre-treatment choline value is obtained. Three of 4 patients who showed histologically confirmed radiation necrosis also had reduced choline levels. The fourth was noted to have increased choline, but a dense fibrous scar affecting the chemical composition was found at later resection.

Figure 3. A (top left), Normal proton MR spectrum of white matter localized in the right parietal cortex of a 23-year-old woman at TE 270 ms, 12 cc voxel (TR 3000 ms, 128 scans). B (bottom left), Proton MRS from the central region of a grade 3/4 astrocytoma at TE 270 ms, 8 cc voxel (TR 1600 ms, 256 scans). C (right), T2-weighted image indicating the tumor distribution and the investigated volumes. Note the increased choline indicating additional cellularity associated with tumor, the decreased NAA associated with neuronal loss, and the new lactate peak associated with increased metabolic activity. (Reprinted with permission from Frahm et al, 1991)

Despite the apparent improvement of MRS over traditional PET studies, in this study glioblastoma multiforme paradoxically had the lowest average normalized choline values. This is likely to be a consequence of the extensive necrosis seen in these tumors and partial volume limitations of the technique. This aberrant finding of the lowest choline value within the most malignant tumor points out a similar limitation in MRS as found in PET. Namely, the most metabolically active tumors often outgrow their blood supply to produce necrotic centers, which falsely decrease the appearance of metabolic activity. The potential to miss a highly malignant lesion with PET and/or MRS suggests caution in using either technique alone as a definitive, independent test for post-radiation evaluations.

Functional Magnetic Resonance Imaging (fMRI)

The most recent advance in neuroradiology to distinguish radiation necrosis from tumor recurrence is dynamic magnetic resonance imaging, also known as functional MRI (fMRI). Previously, fMRI with an endogenous contrast such as the blood oxygenation level of hemoglobin has been used to identify sites of increased neuronal activation during somatosensory, sensorimotor or linguistic tasks. This technique has replicated many of the key findings of PET studies in terms of mapping language and motor functions in the cerebral cortex, establishing its usefulness as a clinical tool20. More recently, the usefulness of fMRI with an exogenous contrast agent such as gadolinium has been demonstrated in brain tumor patients, where echo-planar techniques with contrast-enhanced susceptibility imaging have clearly demonstrated selective sensitivity to the microvasculature of brain tumors. fMRI provides the spatial and temporal resolution needed for measuring regional tissue blood distribution on a volume basis, and can assess dynamically the rate at which contrast leaks past a disrupted blood-brain barrier17. By using long echo times (80 to 100 ms), the large magnetic susceptibility effects of a bolus of gadolinium contrast can be quantified relative to signal loss, and converted to contrast-tissue concentration curves. These concentration curves are displayed as regional cerebral blood volume (rCBV) maps which can be superimposed over traditional MR images, displaying both functional localization and anatomical structures in a single image.

Figure 4. A, MRI cerebral blood volume (rCBV) map in a patient with a grade 3/4 astrocytoma. B, Corresponding FDG-PET uptake image. A good agreement is seen in both normal brain regions and within the tumor, which is located within the right thalamic region. A focal area of intense FDG uptake is seen within the medial portion of the tumor on PET imaging, correlating with the region of highest blood volume on the MRI rCBV map. (Reprinted with permission from Rosen et al, 1993)

Figure 5. A, T1-weighted postcontrast MR image. B, Proton density MR image. C, MRI rCBV map. D, PET study. This image series demonstrates MRI blood volume mapping's ability to depict areas of high microvascularity in the tumor not seen in the conventional post-contrast enhanced or T2-weighted images. In this case, the PET study is reasonably well correlated with the MRI rCBV map. The MRI rCBV map, however, is particularly sensitive for capillary blood volume and relatively insensitive for larger vessels in the brain, which are prominent in the PET study. (Reprinted with permission from Rosen et al, 1993)

Ultrafast echo-planar techniques are particularly sensitive in depicting microvascular changes, and can enable the detection of neovascularization at the capillary level18 (Figure 4). Instead of a traditional MR image which can only identify breakdown of the blood-brain barrier in the form of contrast enhancement, rCBV maps can most accurately identify recurrent tumor by identifying formation of pathologic blood vessels19 (Figure 5). A lack of neovascularity on rCBV would more likely be associated with radiation necrosis, in which new blood vessels usually are not formed. In a study of 55 intracranial tumors, rCBV maps were shown to be closely correlated to tumor grade21. Low-grade gliomas often showed low homogeneous rCBV, whereas high-grade gliomas demonstrated significant heterogeneity and areas of high rCBV. An additional report by the same group further confirmed these findings22, allowing researchers to suggest that fMRI might be used alone as a marker of post-radiation change. A study from the radiation oncology literature examined the role of fMRI in the evaluation and radiotherapeutic treatment planning of patients with malignant glioma, and found correlation between PET and fMRI in 6 of 8 patients23. In the 2 discordant cases, fMRI showed higher rCBV than PET, suggesting that discrepancies with fMRI will at least err toward identifying malignancy and not overlook recurrent tumor. Furthermore, the authors reported two patients in which functional imaging provided key additional information in the planning of radiation therapy, subsequently leading to modification in dose parameters. Therefore, further refinement of this technique may allow for more accurate radiation treatment planning, since its resolution at the capillary level may detect microscopic spread that would not otherwise be included in the traditional radiation field.

CONCLUSION

Distinguishing radiation necrosis from tumor recurrence is a key aspect of post-radiation follow-up in patients with malignant gliomas. The accuracy of this diagnosis determines whether a patient receives the correct therapy of radiation or surgery, versus no further treatment. As radiation necrosis and tumor angiogenesis may both cause breakdown of the blood-brain barrier, they both demonstrate similar enhancement patterns on CT and MRI. Therefore, further imaging techniques are required to distinguish between the two. Despite the usefulness of PET in making this distinction, its clear limitations and relative unavailability make it necessary to search for more accurate, easily available methods. Fortunately, two of the most promising techniques for making this distinction, MR spectroscopy and functional MRI, utilize magnetic resonance technology which is easily available at most hospitals. With continued development, the unique ability of fMRI to provide resolution at the capillary level may eventually lead to significant improvements in both the pre- and post-radiation care of patients with CNS tumors. The fact that MRS, fMRI, as well as conventional MRI, utilize the same instrumentation makes it possible for all three approaches to be implemented together to obtain corroborative and/or complementary data that may improve both sensitivity and specificity of diagnosis.

ACKNOWLEDGMENT

The author wishes to thank Stephen Chan, MD, for his valuable advice and support.

REFERENCES

1. Squire LF, Novelline RA. Neuroradiology. In: Fundamentals of Radiology. Cambridge: Harvard University Press, 1988:312-334.

2. Sheline GE Wara NM, Smith V. Therapeutic irradiation and brain injury. Int J Radiat Oncol Biol Phys 1980;6:1215-1228.

3. Burger DC, Boyko OB. The pathology of central nervous system radiation injury. In: Gatin DH, Leibel SA, Sleline GE, eds. Radiation Injury to the Nervous System. New York: Raven Press, 1991:191-208.

4. Doyle WK, Budinger TF, Valk PE, et al. Differentiation of cerebral radiation necrosis following radiotherapy from tumor recurrence by F-18-FDG and Rb-82 positron emission tomography. J Comput Assist Tomog 1987;11:563-570.

5. Kelly PJ, Daumas-Duport C, Scheithauer BW, et al. Stereotactic histologic correlations of computed tomography and magnetic resonance imaging-defined abnormalities in patients with glial neoplasms. Mayo Clin Proc 1987;62:450-459.

6. Yue NC. Advances in brain tumor imaging. Curr Opin Neurol 1993;6:831-840.

7. DiChiro G, DeLaPaz Rl, Brooks RA, et al. Glucose utilization of cerebral gliomas measured by [18F] fluoro deoxyglucose and positron emission tomography. Neurology 1982;32:1323-1329.

8. Alavi JB, Alavi A, Chawla KJ, et al. Positron emission tomography in patients with glioma: a predictor of prognosis. Cancer 1988;62:1074-1078.

9. Davis WK, Boyko OB, Hoffman JM, et al. [18F] 2-fluoro-2-deoxyglucose-positron emission tomography correlation of gadolinium-enhanced MR imaging of central nervous system neoplasia. AJNR 1993;14:515-523.

10. Valk PE, Budinger TF, Leven VA, et al. PET of malignant cerebral tumors after interstitial brachytherapy. Demonstration of metabolic activity and correlation with clinical outcome. J Neurosurg 1988;19:830-838.

11. Olivero WC, Dulebohn SC, Lister JR. The use of PET in evaluating patients with primary brain tumours: Is it useful? J Neurol Neurosurg Psychiatry 1995;58:250-252.

12. Williams DW. Brain and Its Coverings. In: Chen YM, Pope TL, Ott DJ, eds. Basic Radiology. New York: McGraw-Hill, 1996:395.

13. Shulman RG, Blumire AM, Rothman DL. Nuclear magnetic imaging and spectroscopy of human brain function. Proc Natl Acad Sci USA 1993;90:3127-3133.

14. Cousins JP. Clinical MR spectroscopy: fundamentals, current applications, and future potential. AJNR 1995;164:1337-1347.

15. Frahm J, Bruhn H, Hanike W, et al. Localized proton NMR spectroscopy of brain tumors using short-echo time STEAM sequences. J Comput Assist Tomogr 1991;15:915-922.

16. Fulham MJ, Bizzi A, Dietz MJ, et al. Mapping of brain tumor metabolites with proton MR spectroscopic imaging: Clinical relevance. Radiology 1992;185:675-682.

17. Rosen BR, Aronen HJ, Kwong KK, et al. Advances in clinical neuroimging: Functional MR imaging techniques. Radiographics 1993;13:889-896.

18. Rosen BR, Aronen HJ, Cohen MS, et al. Diffusion and perfusion fast scanning in brain tumors. Neuroimag Clinic North America 1993;3:631-648.

19. Weiskoff R, Davis T. Functional MR imaging of capillary hemodynamics. In: Haucke EM, Potchen EJ, Gottschal KA, Siebert JE, eds. Magnetic resonance angiography: concepts and applications. Chicago: Mosby-Year Book, 1993:473-484.

20. LeBihan D, moderator. Functional magnetic resonance imaging of the brain. Ann Intern Med 1995;122:296-303.

21. Aronen HJ, Gazit IE, Louis DN, et al. Cerebral blood volume maps of gliomas: Comparison with tumor grade and histologic findings. Radiology 1994;191:41-51.

22. Aronen HJ, Glass J, Pardo FS, et al. Echo-planar MR cerebral blood volume mapping of gliomas. Clinical utility. Acta Radiologica 1995;36:520-528.

23. Pardo FS, Aronen HJ, Kennedy D, et al. Functional cerebral imaging in the evaluation and radiotherapeutic treatment planning of patients with malignant glioma. Int J Radiation Oncology Biol Phys 1994;3:663-669.