Malignant Gliomas: MR Imaging Spectrum of Radiation Therapy and Chemotherapy-induced Necrosis of the Brain after Treatment 1

Ashok J. Kumar, MD

Norman E. Leeds, MD

Gregory N. Fuller, MD, PhD

Pamela Van Tassel, MD

Moshe H. Maor, MD

Raymond E. Sawaya, MD

Victor A. Levin, MD

radiology

Radiations, injurious effects, complications of therapeutic radiology, 13.47

Radiology 2000;217:377–384

1_{From the Division of Diagnostic} Im-aging (A.J.K., N.E.L., P.V.T.), and the Departments of Pathology (G.N.F.), Radiation Oncology (M.H.M.), Neuro-surgery (R.E.S.), and Neuro-Oncology (V.A.L.), University of Texas M.D. Anderson Cancer Center, Box 57, 1515 Holcombe Blvd, Houston, TX 77030. From the 1995 RSNA scientific assembly. Received October 20, 1999; revision requested November 19; final revision received March 17, 2000; ac-cepted April 24. V.A.L. and G.N.F. sup-ported in part by NIH grant CA 55261. Address correspondence to A.J.K. (e-mail:akumar@mdanderson.org). Current address:

2_{Department of Radiology, Medical} University of South Carolina, Charles-ton.

©_{RSNA, 2000}

Author contributions:

Guarantors of integrity of entire study, A.J.K., N.E.L., G.N.F., V.A.L.; study con-cepts and design, V.A.L., A.J.K.; defini-tion of intellectual content, V.A.L., N.E.L., A.J.K.; literature research, N.E.L., A.J.K., V.A.L.; clinical studies, all authors; data acquisition and analysis, V.A.L., N.E.L., A.J.K.; manuscript preparation and editing, A.J.K., N.E.L., G.N.F.; manuscript review, all authors.

Malignant Gliomas: MR

Imaging Spectrum of

Radiation Therapy– and

Chemotherapy-induced

Necrosis of the Brain

after Treatment

1

PURPOSE:To describe both the common and less frequently encountered mag-netic resonance (MR) imaging features of radiation therapy– and chemotherapy-induced brain injury, with particular emphasis on radiation necrosis.

MATERIALS AND METHODS:A cohort of 148 adult patients underwent surgical resection of malignant brain (glial) tumors and were subsequently entered into a research protocol that consisted of accelerated radiation therapy with carboplatin followed by chemotherapy with procarbazine, lomustine, and vincristine. Patients typically underwent sequential MR imaging at 6 – 8-week intervals during the 1st year and at 3– 6-month intervals during subsequent years. In all patients, his-topathologic confirmation of lesion composition was performed by board-certified neuropathologists.

RESULTS:The patients exhibited different types of MR imaging– detected abnor-malities of the brain: pure radiation necrosis in 20 patients, a mixture of predomi-nantly radiation necrosis with limited recurrent and/or residual tumor (less than 20% of resected tissue) in 16 patients, radiation necrosis of the cranial nerves and/or their pathways in two patients, radiation-induced enhancement of the white matter in 52 patients, and radiation-induced enhancement of the cortex in nine patients.

CONCLUSION: The frequent diagnostic dilemma of recurrent neoplasm versus radiation necrosis is addressed in this study through a description of the varying spatial and temporal patterns of radiation necrosis at MR imaging.

During the past decade, the treatment of malignant gliomas has benefitted from several advances, including early detection of tumors by using magnetic resonance (MR) imaging, improvements in neurosurgical instrumentation that have facilitated more precise radical resection of neoplasms, advances in the delivery of radiation doses to tumors, and new chemotherapeutic protocols. However, the sequential MR images obtained in these pa-tients have demonstrated a variety of radiation therapy– and chemotherapy-induced changes in the brain. Radiation necrosis is the most substantial and most severe form of radiation-induced injury with therapeutic implications. The diagnosis of radiation necro-sis at imaging has been challenging, primarily because the pattern of abnormal enhance-ment closely mimics that of recurrent brain tumor. The purpose of this retrospective study was to review our experience with the MR imaging spectrum of radiation necrosis. MATERIALS AND METHODS

One hundred forty-eight patients (82 men, 66 women; mean age, 41 years; age range, 18 –75 years) with malignant gliomas—92 with glioblastoma multiforme, 40 with

anaplas-tic astrocytoma, nine with anaplasanaplas-tic mixed glioma, and seven with anaplastic oligodendroglioma—were entered into research protocol DM88-113 after they gave institutional review board–approved informed consent and after undergoing either total or subtotal resection of their tumors (1). After surgery, the patients re-ceived accelerated fractionation radia-tion therapy and concomitant carbopla-tin-based chemotherapy. The accelerated radiation therapy consisted of three doses of 1.9 –2.0 Gy each per day, with a 4-hour interval between fractions. Carbo-platin was given intravenously for 2 hours at a dose of 33 mg/m2_{, 1.75 hours} before each radiation treatment. This reg-imen was administered with a 6-MV lin-ear accelerator for 5 consecutive days. Af-ter a 2-week break, a second course of treatment identical to the first was given so that a total dose of 57– 60 Gy was given; the majority of patients received 57 Gy following an amendment to the protocol.

The radiation portal was generally lim-ited to a 2–3-cm margin around the en-hancing tumor and the nonenen-hancing abnormal tumoral signal intensity (not edema), if present, at T2-weighted imag-ing. In tumors that were closer to the midline—for example, corpus callosal tu-mors—partially opposing portals were used and resulted in margins wider than 3 cm in the lateral dimension. Postradia-tion chemotherapy consisted of procarba-zine, lomustine, and vincristine, beginning within 4 weeks after the completion of ra-diation therapy. The administration of procarbazine, lomustine, and vincristine was repeated at 6-week intervals for 1 year or until tumor progression.

The preoperative and postoperative follow-up MR images of the brain ob-tained in the 148 patients were reviewed by two authors (A.J.K., N.E.L.), and the findings were tabulated by means of con-sensus. The information gathered from the MR images consisted of (a) the en-hancing or nonenen-hancing features of pri-mary neoplasm, (b) the appearance and location of recurrent enhancing lesions, single or multiple, (c) the proximity of the recurrent lesion to the primary tu-mor, and (d) the interval between the end of radiation therapy and the histopatho-logic confirmation of radiation necrosis. The MR images of the brain consisted of standard nonenhanced and contrast ma-terial– enhanced studies that were ob-tained both at our institution and at out-side imaging centers.

At our institution, the MR images of the brain, which were obtained by using

a 1.5-T unit (Signa Horizon 5X; GE Med-ical Systems, Milwaukee, Wis), consisted of transverse T2-weighted (2,000 –3,000/ 80 –100 [repetition time msec/echo time msec], one signal acquired) and interme-diate-weighted (2,000 –3,000/20 –30, one signal acquired) images and pre- and postcontrast transverse, T1-weighted, spin-echo images (500 – 600/8 –16, two signals acquired). The postcontrast transverse and coronal T1-weighted images were obtained by administering gadopentetate dimeglumine (Magnevist; Berlex Labora-tories, Wayne, NJ) intravenously at a dose of 0.1 mmol per kilogram of body weight. Occasionally, sagittal postcon-trast T1-weighted images of the brain also were obtained as needed.

In the patients in the DM88-113 pro-tocol, postsurgical follow-up MR images were obtained either within 72 hours or 2– 4 weeks after surgery. Follow-up MR images were obtained within 2 weeks after completing radiation therapy and at 6 – 8-week intervals during the pro-carbazine, lomustine, and vincristine treatment period. During year 2 after treatment, follow-up MR images were ob-tained at 3-month intervals; during year 3, at 4-month intervals; and during year 4, at 6-month intervals. If clinical deteri-oration occurred, MR images were ob-tained as needed.

Histopathologic examination of 36 surgically resected tissue specimens— nine by means of stereotactic biopsy and 27 by means of open resection—was per-formed by board-certified neuropatholo-gists (one of which was G.N.F.). The au-topsy results in nine patients also were available.

RESULTS

Therapy-induced necrosis in the brain, which was proved at histopathologic ex-amination, was found in 22 of the 56 patients with anaplastic gliomas and in 14 of the 92 patients with glioblastomas under protocol DM88-113. The necrosis of the brain was attributed to the com-bined effects of radiation therapy and chemotherapy. Isodose curves revealed that all of the radiation-induced necroses occurred within the radiation portal. Analysis of the long-T2–weighted images was not found to be helpful in differen-tiating recurrent tumor from radiation necrosis.

Twenty of the 36 surgical specimens had histopathologic evidence of radia-tion necrosis with no evidence of tumor, and 16 had evidence of a mixture of

ex-tensive necrosis with residual and/or re-current tumor (less than 20% of resected tissue). The common and uncommon features of radiation necrosis included ra-diation necrosis simulating a recurrent glioma at the site of the surgical cavity (Fig 1), distant to the site of the primary tumor on the ipsilateral side (Fig 2), or on the contralateral side (Fig 3), and multi-ple radiation-induced necrotic masses simulating multiple metastases (Fig 4a) and/or multiple sclerosis (Fig 4b).

The MR imaging features commonly seen in radiation necrosis are a soap bubble–like interior (Fig 4a) and a Swiss cheese–like interior (Fig 5). The MR im-aging characteristics of radiation necrosis in both groups—that is, in the patients with pure radiation necrosis and in those with predominantly radiation necrosis— are outlined in the Table. In the 20 pa-tients with pure radiation necrosis, the primary tumor was nonenhancing in 10 anaplastic gliomas and enhancing in 10 tumors—three anaplastic gliomas and seven glioblastomas. In the 16 patients with necrosis intermingled with tumor, the primary tumor was enhancing in three anaplastic gliomas and seven glio-blastomas and nonenhancing in six ana-plastic gliomas. Infratentorial radiation-induced necrosis following radiation therapy for parietal lobe glioblastoma is illustrated in Figure 6. The intervals from completion of radiation therapy to his-topathologic confirmation of necrotic le-sions in both groups are illustrated in Figure 7.

Radiation-induced enhancement in the brain without radiation necrosis was observed in the white matter (Fig 8) in 52 patients and in the cortex (Fig 9) in nine patients. Radiation-induced necrosis of the optic pathway (Fig 10) was seen in one patient, and radiation-induced ne-crosis of the auditory pathway (Fig 11) was seen in one other patient.

DISCUSSION

Pathophysiology of Radiation-induced Damage to the Brain

Currently, a complete understanding of the pathophysiology of radiation ther-apy– and chemotherapy-induced injury to the central nervous system is lacking. Relevant variables include total radiation dose, radiation field size, radiation frac-tion size, number and frequency of radi-ation doses, combinradi-ation of radiradi-ation therapy and chemotherapy, duration of survival, and age of the patient at treat-ment (2). Understanding the importance

of specific MR imaging findings in this patient population is aided by a familiar-ity with the current concepts of the pathophysiology of radiation-induced damage to the central nervous system, which are reviewed in the following text. The postulated mechanisms that may contribute to radiation-induced neurotox-icity include (a) vascular injury, (b) glial and white matter damage, (c) effects on the

fibrinolytic enzyme system, and (d) im-mune mechanisms.

Vascular injury.—In acute radiation-in-duced injury, transient vasodilatation oc-curs with variable changes in capillary permeability that sometimes manifest as vasogenic edema (3,4). Changes in capil-lary permeability, in response to radia-tion therapy to the central nervous sys-tem, have been quantitated and found to

entail specific molecular-size variability (5,6). In chronic radiation-induced in-jury, vascular endothelial damage takes place (4). Fike et al (7) demonstrated in animal studies that vascular abnormali-ties occur before the development of pa-renchymal changes in the brain. The pathologic findings consist of endothe-lial damage, vascular ectasia, and telangi-ectasia, all of which result in increased Figure 1. Radiation necrosis simulating a recurrent glioma in a 44-year-old woman who underwent resection of a nonenhancing anaplastic mixed

glioma in the left anterior temporal lobe followed by accelerated radiation therapy (60 Gy) and chemotherapy.(a)Postcontrast coronal T1-weighted

spin-echo image (700/16) obtained 17 months after radiation therapy shows ringlike areas of abnormal enhancement (solid arrows) have developed

around the surgical cavity. A left paraventricular enhancing lesion (open arrow) characteristic of posttreatment change also is noted.(b)Histologic

section reveals a characteristic finding in radiation necrosis—a blood vessel with fibrinoid necrosis of the wall (N) and a narrowed lumen (arrow).

(Hematoxylin-eosin stain; original magnification, ⫻400.)(c) High-power photomicrograph reveals radiation-induced telangiectasia (arrows).

(Hematoxylin-eosin stain; original magnification,⫻200.)

Figure 2. Evolution of radiation necrosis. Radiation necrosis with tumorlike growth arising distant (anterior and inferior) to the site of the primary tumor in a 22-year-old man who had been treated with surgery, accelerated radiation therapy (57 Gy), and chemotherapy for a

nonenhancing anaplastic astrocytoma in the right parietal lobe.(a)Postcontrast transverse

T1-weighted spin-echo image (600/8) obtained 16 months after completion of radiation therapy shows small enhancing lesions in the right superficial temporal lobe (arrow) and

paraventricular region (arrowhead) of the deep temporal lobe.(b)Postcontrast transverse

T1-weighted spin-echo image (600/8) obtained 2 months 15 days later demonstrates a rapid

increase in the size of the lesions (solid arrows) ina, as well as new enhancing lesions (open

arrows). Rapid growth of the paraventricular mass was associated with surrounding edema

(E) and mass effect. Two weeks later, the patient underwent partial excision of the mass, and

histopathologic analysis revealed foci of necrosis, gliosis, and scattered atypical cells but no evidence of recurrent tumor.

Figure 3. Radiation necrosis of the contralat-eral cerebral hemisphere simulating a multi-centric tumor with corpus callosal involve-ment in a 30-year-old woman who underwent resection of a glioblastoma in the left frontal lobe followed by accelerated radiation therapy and chemotherapy. On the postcontrast coro-nal T1-weighted spin-echo image (500/11), the thin black arrows point to the craniotomy site. Thirteen months after radiation therapy, the patient developed an irregularly enhancing mass in the right frontal lobe white matter (large white arrow), cortex (small white arrow),

and corpus callosum (C), with surrounding

edema (E) and subfalcine herniation.

Stereo-tactic biopsy of the enhancing regions revealed radiation necrosis. The open arrows point to fingerlike contrast-enhanced white matter pro-jections toward the arcuate fibers.

capillary permeability at the site of vas-cular injury with resultant cytotoxic and vasogenic edema (4). Progressive vascular changes include vessel wall thickening caused by hyalinization, with conse-quent thrombosis, infarction, and necro-sis (4). The results of these studies provide support for the concept that vascular in-jury has a pivotal role in the development of radiation-induced neurotoxic effects in the brain. This hypothesis, however, has not been proved and is not universally ac-cepted. One alternative concept is that cy-tokine release may stimulate new vessel formation, which contributes to capillary leakage.

Glial and white matter damage.—It has been shown that oligodendrocytes are extremely sensitive to radiation and that their destruction is associated with radio-logic evidence of the demyelination that ensues (4,8). Although neurons are rela-tively insensitive to radiation, there is sufficient loss of cellular components, primarily in white matter, to account for the observed reduction in brain volume that accompanies central nervous system radiation toxicity. Documented meta-bolic changes that occur in irradiated cells—for example, reduced glycolysis— are associated with a decrease in glucose consumption, which correlates with the decrease in glucose and oxygen

utiliza-tion observed on the positron emission tomographic scans obtained in patients with radiation-induced central nervous system necrosis (9).

Effects on the fibrinolytic enzyme sys-tem.—Sawaya (10) examined necrotic brain tissue samples obtained in patients with radiation-induced necrosis and found MR Imaging Characteristics of Radiation Necrosis of the Brain in 36 Patients

Lesion Characteristics Pure Radiation Necrosis (n⫽20) Predominantly Radiation Necrosis (n⫽16)* Solitary Lesion† _{8 11} Multiple Lesions† _{12 5} Enhancement pattern‡ Solid lesion 10 5 Ring lesion 31 17

Coalescent central lesions 14 13

“Soap bubble”–like interior 13 4

“Swiss cheese”–like interior 4 0

Location‡

Corpus callosum 6 3

Periventricular white matter 8 0

Central white

matter-coritcomedullary junction 22 18

Brain stem 3 0

Cerebellar hemisphere 2 0

Basal ganglia 0 1

Proximity to original tumor‡

Within the tumor site 8 7

Edge of treated tumor 11 11

Several cm from tumor

Ipsilateral 14 4

Contralateral 8 0

* Predominantly radiation necrosis with limited recurrent and/or residual tumor. †_{Data are the number of patients.}

‡_{Data are the number of lesions, based on a total of 41 (pure radiation necrosis) or 22} (predominantly radiation necrosis) enhancing lesions.

Figure 4. Autopsy-proved radiation necrosis within both cerebral hemispheres simulating mul-tiple metastases and/or tumefactive mulmul-tiple sclerosis in a 30-year-old man who underwent subtotal resection of a multicentric, nonenhancing anaplastic astrocytoma in the right parietal lobe and left anterior temporal lobe followed by whole-brain accelerated radiation therapy (60

Gy) and chemotherapy.(a)Postcontrast transverse T1-weighted spin-echo image (500/17)

ob-tained 17 months after radiation therapy shows multiple enhancing masses (solid arrows) within the frontal lobe white matter simulating metastases. Central soap bubble–like areas of necrosis (open arrow) were seen within the masses. Autopsy revealed no evidence of tumor, and the histopathologic findings showed extensive areas of radiation necrosis that corresponded to the

multiple enhancing lesions identified at MR imaging. (b) Postcontrast sagittal T1-weighted

spin-echo image (500/17) shows multiple periventricular enhancing lesions involving the adja-cent corpus callosum (arrows) and the temporal horn (arrowheads) that are distributed along the body of the lateral ventricle and simulating tumefactive multiple sclerosis.

Figure 5. Postcontrast coronal T1-weighted follow-up spin-echo image (600/16) ob-tained 3 years 7 months after radiation ther-apy in a 36-year-old man who underwent surgery, accelerated radiation therapy (60 Gy), and chemotherapy for a right frontal lobe nonenhancing anaplastic astrocytoma shows radiation necrosis of the bilateral frontal lobe white matter (white arrows) with a Swiss cheese pattern of interior (straight black arrows) and secondary atrophy of the brain. Arcuate fibers (curved arrows) of the white matter also were affected. Radiation-induced necrosis was identified at computed tomography (CT) 17 months after therapy, and the diagnosis was confirmed by using serial MR images. The patient died 5 years after the initial diagnosis.

an absence of tissue plasminogen activa-tor and an excess of urokinase plasmino-gen activator. These enzymes are mem-bers of a complex fibrinolytic pathway that has potent variable effects on blood vessels and brain tissue. Tissue plasmin-ogen activator affects fibrinplasmin-ogen during blood clotting, whereas urokinase plas-minogen activator is active in extracellu-lar proteolysis. An increase in urokinase plasminogen activator with a concomi-tant decrease in tissue plasminogen acti-vator may contribute to cytotoxic edema and tissue necrosis.

Immune mechanisms.—An intriguing concept, the validity of which is cur-rently unknown, is the possibility of an immune response in the host secondary to tissue damage caused by radiation and chemotherapy. The potential role of au-toimmune vasculitis in the response of the central nervous system to radiation-induced damage needs further investiga-tion (8).

Histopathologic Features

The end point of radiation injury is necrosis that may develop months to

years after the completion of radiation therapy and is frequently irreversible and often progressive (4,11). Radiation necro-sis occurs most commonly around blood vessels within white matter (4,11). The most common histopathologic features are fibrinoid necrosis of blood vessel walls, with surrounding perivascular pa-renchymal coagulative necrosis (Fig 1). Extension and confluence of multiple perivascular foci of necrosis result in large serpiginous or “geographic” zones of parenchymal necrosis (Fig 1), which are followed by the deposition of mineral salts (ie, dystrophic calcification). An ad-ditional vascular lesion that is often ob-served consists of clusters of abnormally dilated, thin-walled telangiectasias (Fig 1c). Late vascular changes include vessel wall thickening caused by hyalinization, with resultant luminal narrowing (Fig 1b). White matter changes include focal and diffuse demyelination.

Radiologic Findings

Several articles (12–26) on the imaging findings of adverse effects of therapeutic radiation to the brain have been pub-lished. Van Tassel et al (23) described MR findings of brain injury after accelerated fractionation radiation therapy com-bined with carboplatin chemotherapy for the treatment of malignant gliomas. The present study is an extension and expansion of that work, with emphasis on both the common and uncommon features of radiation necrosis of the brain. Radiation-induced necrosis of the white matter.—Radiation-induced necrosis is the end result of perivascular coagulative necrosis affecting the white matter. As expected, radiation necrosis occurs most commonly at the site of maximum radi-ation delivery, that is, in the immediate vicinity of the tumor site and surround-ing the surgical cavity of a partially or Figure 6. Infratentorial necrosis following

ra-diation therapy in a 21-year-old woman who underwent surgery, accelerated radiation ther-apy (60 Gy), and chemotherther-apy for a left pari-etal lobe glioblastoma multiforme. On the postcontrast coronal T1-weighted spin-echo image (500/11), the open arrow points to a large surgical cavity with no residual tumor 13 months after resection. The solid arrows point to solid areas of abnormal enhancement in-volving the white matter in both cerebellar hemispheres. This abnormal enhancement was presumed to represent radiation necrosis; however, this was not verified with his-topathologic analysis. Six months later, the patient died of recurrent parietoccipital lobe

glioblastoma. Figure 7._{pletion of radiation therapy to histopathologic confirmation of treatment-induced necrosis.}Chart (top) and plot graph (bottom) illustrate the intervals, in months, from com-AG⫽anaplastic glioma,GBM⫽glioblastoma multiforme. Top: The data in the middle and right columns are numbers of patients. Bottom: The four columns represent box plots of actual data (⫾

SD). The median values are designated by the horizontal line bisecting each box, and the ranges are designated by the upper and lower bars outside the boxes.Necrosisrefers to pure radiation necrosis, andMixrefers to predominantly necrosis with limited residual and/or recurrent tumor.

totally resected tumor (Fig 1). Radiation necrosis can closely resemble recurrent tumor at MR imaging or CT because of the following shared characteristics: (a) or-igin at or close to the oror-iginal tumor site, (b) contrast enhancement, (c) growth over time, (d) edema, and (e) exertion of mass effect. With respect to the MR imaging characteristics of radiation necrosis, most lesions consist of an enhancing mass with a central area of necrosis. The contrast en-hancement of these lesions is secondary to radiation-induced endothelial damage, which leads to the breakdown of the blood-brain barrier. The use of platinum-based chemotherapy drugs, such as cispla-tin and carboplacispla-tin, combined with radia-tion therapy may contribute to the development of radiation-induced necrosis (27–30). On T2-weighted images, the solid portion of the radiation-induced necrotic mass has low signal intensity, and the cen-tral necrotic component shows increased signal intensity.

In addition to the most common pat-tern of radiation necrosis, which consists of a single lesion arising at the site of the original primary tumor, other less com-mon patterns also may be observed. Ex-amples include (a) multiple lesions, (b) lesions in the contralateral hemi-sphere, (c) lesions arising remotely from a primary cerebral site—for example, in the cerebellum (Fig 6) or brain stem, and (d) subependymal lesions. Each of these rarer patterns of radiation necrosis can be suggestive of a different pathologic pro-cess. For example, radiation necrosis oc-curring at the site of the resected primary tumor can be easily mistaken for recur-rent tumor (Fig 1). Necrosis occurring dis-tant to the primary site of tumor may mimic multifocal glioma (Fig 2). In Fig-ure 3, the radiation necrosis around the periventricular white matter of the con-tralateral hemisphere simulating multi-centric glioma has tumorlike associated edema and mass effect. Marks et al (2) described radiation necrosis arising con-tralaterally to the primary tumor site in one patient following whole-brain irradi-ation. In our series, we observed this find-ing in eight patients.

Multiple lesions can resemble multiple metastases (Fig 4a). The predilection for periventricular white matter involvement in radiation necrosis may, on rare occa-sion, mimic tumefactive multiple sclero-sis (Fig 4b). The radiation necrosclero-sis that develops in the periventricular white matter may be explained by the fact that this neuroanatomic region has a rela-tively poor blood supply from long med-ullary arteries that lack collateral vessels

and is therefore vulnerable to ischemic effects produced by postradiation vascu-lopathy (31,32). Therapy-induced ne-crotic lesions can also spread subependy-mally and mimic subependymal tumor spread (Fig 2b).

As radiation necrosis progresses with tumorlike growth, it can lead to severe shrinkage of the white matter and cortex and result in focal brain atrophy (Fig 5). The arcuate fibers in white matter are relatively resistant to radiation necrosis and are usually involved late in the dis-ease process. An additional variant of ra-diation necrosis that warrants brief atten-tion is the Swiss cheese pattern. This pattern can be visualized as a result of diffuse necrosis affecting the white mat-ter and adjacent cortex. The Swiss cheese pattern reflects diffuse enhancements at the margins affecting the cortex and white matter, with intermixed foci of ne-crosis. Compared with lesions with the soap bubble pattern, Swiss cheese lesions are larger, more variable in size, and more diffuse (Fig 5).

The incidence of radiation necrosis af-ter conventional therapy ranges from 5% to 24%, with higher rates at autopsy (2,33). In our series of 148 patients with treated malignant gliomas, pure radia-tion necrosis was observed in 20 (13.5%)

patients, and a mixture of predominantly radiation necrosis intermingled with lim-ited residual and/or recurrent tumor (less than 20% of resected tissue) was found in 16 (10.8%). It is important to remember that radiation-induced necrosis is a dy-namic pathophysiologic process with several possible clinical outcomes. Al-though continued growth with atten-Figure 8. Spontaneous regression of

radia-tion-induced multifocal enhancement of the white matter that gradually disappeared within 29 months. Follow-up postcontrast transverse T1-weighted spin-echo image (500/11) obtained in a 41-year-old man who underwent surgery, accelerated radiation therapy, and chemother-apy for a nonenhancing anaplastic astrocytoma involving the left parietal lobe shows multinod-ular foci of radiation-induced enhancement (ar-rows) within the white matter.

Figure 9. Radiation-induced damage to the cortex simulating subacute infarction. Post-contrast transverse T1-weighted spin-echo im-age (500/11) obtained 17 months after therapy in a 34-year-old man who had received accel-erated radiation therapy (57 Gy) and chemo-therapy for an anaplastic astrocytoma involv-ing the left cinvolv-ingulate gyrus shows an enhancinvolv-ing area (arrow) in the right insular cortex. This enhancing area was present on the follow-up MR image (not shown) obtained 1 year later.

Figure 10. Postcontrast transverse T1-weighted fat-suppressed spin-echo image (500/11) ob-tained in a 26-year-old man after surgery, 9 months after accelerated radiation therapy (55.5 Gy) and chemotherapy for a frontal lobe ana-plastic astrocytoma, shows radiation necrosis in-volving the optic nerves (open arrows), chiasm (arrowheads), and tracts (solid arrows). The ne-crosis was confirmed at autopsy.

dant cytotoxic edema and mass effect is commonly seen, lethal progression is not inevitable in all cases. Some lesions will stabilize, and others will regress in size. Surgery may be required to reduce the mass effect and edema, and despite advancements in special imaging tech-niques, surgery may be needed to estab-lish an accurate histopathologic diagno-sis. Finally, despite the use of total gross resection, which includes removal of the necrotic tissue, radiation necrosis recur-rence will be observed in some cases. It is anticipated that current efforts to further refine investigational radiologic tech-niques will lead to less ambiguous diag-noses of radiation-induced necrosis in the near future.

Radiation-induced enhancement of the white matter and cortex.—Radiation-in-duced enhancement, a milder form of brain injury, usually occurs within the white matter, can be single or multiple, and appears in varying sizes. This en-hancement mostly affects the white mat-ter at variable distances from the primary tumor site (Fig 8). The contralateral cere-bral hemisphere or cerebellar white mat-ter is affected less often. The patmat-tern of enhancement can be nodular (Fig 8), lin-ear (Fig 1), or curvilinlin-ear. These lesions are frequently seen at MR imaging and do not warrant biopsy if they remain sta-ble (Fig 9) or regress (Fig 8) in size. How-ever, progression to radiation necrosis should be suspected, and if they have an increase in size, accompanied by edema and mass effect (Fig 2). In such instances, surgery is recommended. Radiation-in-duced damage of this type can also occur within the cortex. For example, cortical gyral enhancement may simulate sub-acute infarction (Fig 9).

Radiation-induced necrosis of the cranial nerves.—Radiation-induced cranial neu-ropathy is relatively rare. Radiation-in-duced optic neuropathy (Fig 10) is likely to occur when focused radiation is deliv-ered to tumors in the perioptic regions (34). The pathologic process of this neu-ropathy consists of optic pathway vascu-lopathy with secondary hemorrhage, re-active gliosis, necrosis, and atrophy. A similar process can occur around other cranial nerves, as demonstrated in a pa-tient in our series with necrosis at and around the lateral lemniscus of the cen-tral auditory pathway (Fig 11).

In summary, the frequent diagnostic dilemma of recurrent neoplasm versus ra-diation necrosis is addressed in this study through a description of the varying spa-tial and temporal patterns of radiation

necrosis that may be encountered at MR imaging:

1. If the tumor was a nonenhancing lesion before surgery and enhancing foci subsequently developed within and cir-cumscribing the tumor, the enhancing lesions often do not represent progres-sion to a higher grade, but rather they are often the result of radiation necrosis.

2. If an enhancing focus develops at a distance from the primary glioma, radia-tion necrosis should be suspected; multi-ple enhancing areas do not necessarily represent multifocal glioma.

3. If an enhancing focus develops in the periventricular white matter, partic-ularly capping the ventricles or within the corpus callosum, radiation necrosis should be suspected. The periventricular white matter is among the areas most susceptible to radiation necrosis.

4. Radiation necrosis should be sus-pected if a new enhancing lesion exhibits a soap bubble or Swiss cheese pattern.

The key point in any of these scenarios is that radiation necrosis should be in-cluded high in the differential diagnosis. It is important to recognize the spectrum of MR imaging features of radiation ne-crosis, because this tissue damage can mimic other pathologic processes. Acknowledgments: We thank Marites A. Tre-vino and Kari Maloney for help in the prepa-ration of this manuscript and Don Norwood, BS, for editorial assistance.

References

1. Levin VA, Moar MH, Thall PF, et al. Phase

II study of accelerated fractionation radi-ation therapy with carboplatin followed by vincristine chemotherapy for the treatment of glioblastoma multiforme. Int J Radiat Oncol Biol Phys 1995; 33: 357–364.

2. Marks JE, Baglan RJ, Prassad SC, Blank

WF. Cerebral radionecrosis: incidence and risk in relation to dose, time, frac-tionation and volume. Int J Radiat Oncol Biol Phys 1981; 7:243–252.

3. Husain MM, Garcia JH. Cerebral “radiation

necrosis”: vascular and glial features. Acta Neuropathologica 1976; 36:381–385.

4. Burger PC, Boyko OB. The pathology of

central nervous system radiation injury. In: Gutin PH, Leibel SA, Sheline GE, eds. Radiation injury to the nervous system. New York, NY: Raven, 1991; 191–208.

5. Levin VA, Edwards MS, Byrd A.

Quantita-tive observations of the acute effects of X-irradiation on brain capillary perme-ability: part I. Int J Radiat Oncol Biol Phys 1979; 5:1627–1631.

6. Edwards MS, Levin VA, Byrd A.

Quantita-tive observations of the subacute effects of x irradiation on brain capillary perme-ability: part II. Int J Radiat Oncol Biol Phys 1979; 5:1633–1635.

7. Fike JR, Sheline GE, Cann CE, Davis RL.

Radiation necrosis. Prog Exp Tumor Res 1984; 28:136 –151.

Figure 11. Radiation necrosis of the pons in-volving the white matter tracts, including the region of the auditory pathway lateral lemnis-cus, in a 44-year-old woman who had under-gone resection of a left temporal lobe ana-plastic astrocytoma followed by accelerated radiation therapy (60 Gy) and chemotherapy.

(a)Postcontrast transverse T1-weighted

spin-echo image (700/16) shows radiation-induced necrosis involving the lateral aspect of the pons (solid arrow). The open arrow points to an enhancing area posterior to the surgical cavity in the temporal lobe that was his-topathologically proved to be radiation

necro-sis.(b)Whole-mount histologic section of the

left pons inashows pallor of the fiber tracts

and nuclei (arrows). The principal fiber tract of the auditory pathways, the lateral lemniscus, is included within the area of demyelination. (Hematoxylin-eosin stain; original

8. Castel JC, Caille JM. Imaging of irradiated brain tumours: value of magnetic reso-nance imaging. J Neuroradiol 1989; 16: 81–132.

9. Di Chiro G, Oldfield E, Wright DC, et al.

Cerebral necrosis after radiotherapy and/or intraarterial chemotherapy for brain tu-mors: PET and neuropathologic studies. AJR Am J Roentgenol 1988; 150:189 –197.

10. Sawaya R. The fibrinolytic enzymes in the

biology of brain tumors. In: Sawaya MD, ed. Fibrinolysis and the central nervous system. Philadelphia, Pa: Hanley & Bel-fus, 1990; 106 –126.

11. Lampert PW, Davis RL. Delayed effects of

radiation on the human central nervous system: “early” and “late” delayed reac-tions. Neurology 1964; 14:912–917.

12. Mikhael MA. Radiation necrosis of the

brain: correlation between patterns on computer tomography and dose of radia-tion. J Comput Assist Tomogr 1979; 3:241–249.

13. Kingsley DP, Kendall BE. CT of the

ad-verse effects of therapeutic radiation of the central nervous system. AJNR Am J Neuroradiol 1981; 2:453– 460.

14. Safdari H, Boluix B, Gros C. Multifocal

brain radionecrosis masquerading as tu-mor dissemination. Surg Neurol 1984; 21: 35– 41.

15. Curnes JT, Laster DW, Ball MR, Moody

DM, Witcofski RL. MRI of radiation in-jury to the brain. AJR Am J Roentgenol 1986; 147:119 –124.

16. Dooms GC, Hecht S, Brant-Zawadzki M,

Berthiaume Y, Norman D, Newton TH. Brain radiation lesions: MR imaging. Ra-diology 1986; 158:149 –155.

17. Curran WJ, Hecht-Leavitt C, Schut L,

Zimmerman RA, Nelson DF. Magnetic resonance imaging of cranial radiation le-sions. Int J Radiat Oncol Biol Phys 1987; 13:1093–1098.

18. Constine LS, Konski A, Ekholm S,

Mc-Donald S, Rubin P. Adverse effects of brain irradiation correlated with MR and CT imaging. Int J Radiat Oncol Biol Phys 1988; 15:319 –330.

19. Valk PE, Dillon WP. Radiation injury of

the brain. AJNR Am J Neuroradiol 1991; 12:45– 62.

20. Ball WS Jr, Prenger EC, Ballard ET.

Neuro-toxicity of radio/chemotherapy in chil-dren: pathologic and MR correlation. AJNR Am J Neuroradiol 1992; 13:761–776.

21. Nishimura R, Takahashi M, Morishita S,

Sumi M, Uozumi H, Sakamoto Y. MR Gd-DTPA enhancement of radiation brain in-jury. Radiat Med 1992; 10:109 –116.

22. Packer RJ, Zimmerman RA, Kaplan A, et

al. Early cystic/necrotic changes after hy-perfractionated radiation therapy in chil-dren with brain stem gliomas: data from the Childrens Cancer Group. Cancer 1993; 71:2666 –2674.

23. Van Tassel P, Bruner JM, Maor MH, et al.

MR of toxic effects of accelerated fraction-ation radifraction-ation therapy and carboplatin chemotherapy for malignant gliomas. AJNR Am J Neuroradiol 1995; 16:715–726.

24. Rabin BM, Meyer JR, Berlin JW,

Mary-mount MH, Palka PS, Russell EJ. Radia-tion-induced changes in the central nervous system and head and neck. Ra-dioGraphics 1996; 16:1055–1072.

25. Santoni R, Liebsch N, Finkelstein DM, et

al. Temporal lobe (TL) damage following surgery and high-dose photon and

pro-ton irradiation in 96 patients affected by chordomas and chondrosarcomas of the base of the skull. Int J Radiat Oncol Biol Phys 1998; 41:59 – 68.

26. Wong ET. Recurrent cystic radiation

ne-crosis of the brain. Oncol Rep 1998; 5:685– 687.

27. Douple EB, Richmond RC, O’Hara JA,

Coughlin CT. Carboplatin as a potentia-tor of radiation therapy. Cancer Treat Rev 1985; 12:111–124.

28. Nias AH. Radiation and platinum drug

interaction. Int J Radiat Biol 1985; 48: 297–314.

29. Fu KK, Rayner PA, Lam KN. Modification

of the effects of continuous low dose rate irradiation by concurrent chemotherapy infusion. Int J Radiat Oncol Biol Phys 1984; 10:1473–1478.

30. Thames HD Jr, Peters LJ, Withers HR,

Fletcher GH. Accelerated fractionation vs hyperfractionation: rationales for several treatments per day. Int J Radiat Oncol Biol Phys 1983; 9:127–138.

31. Moody DM, Bell MA, Challa VR. Features

of the cerebral vascular pattern that predict vulnerability to perfusion or oxygenation deficiency: an anatomic study. AJNR Am J Neuroradiol 1990;11:431– 439.

32. Nelson MD Jr, Gonzalez-Gomez I, Gilles

FH. The search for human telencephalic ventriculofugal arteries. AJNR Am J Neu-roradiol 1991; 12:215–222.

33. Burger PC, Mahley MS Jr, Dudka L, Vogel

FS. The morphologic effects of radiation administered therapeutically for intracra-nial gliomas: a postmortem study of 25 cases. Cancer 1979; 44:1256 –1272.

34. Hudgins PA, Newman NH, Dillon WP,

Hoffman JC Jr. Radiation-induced optic neuropathy: characteristic appearances on gadolinium-enhanced MR. AJNR Am J Neuroradiol 1992; 13:235–238.