Magnetic resonance imaging in the clinical diagnosis of late temporal lobe necrosis following radiotherapy for nasopharyngeal carcinoma

Clinical Radiology (1990) 42, 24-31

Magnetic Resonance Imaging in the Clinical Diagnosis of Late Temporal Lobe Necrosis following Radiotherapy for Nasopharyngeal Carcinoma A. W. M. LEE, L. O. C. CHENG*, S. H. NGt, V. K. C. TSE, S. K. O, G. K. H. AU and Y. F. POON

Medical and Health Department Institute of Radiology and Oncology, Queen Elizabeth Hospital, *Magnetic Imaging Centre, Canossa Hospital, and tHonorary Consultant Neurologist, Baptist Hospital, Hong Kong

Magnetic resonance imaging (MRI) and computed tomography (CT) were performed on 13 patients with clinical features of late temporal lobe damage following radical radiotherapy for nasopharyngeal carcinoma, and their results compared. MRI was undoubtedly superior in sensitivity-revealing areas of prolonged T2 relaxation time in the inferior portions of the temporal lobes in all patients, while CT failed to show any gross abnormalities in seven. In patients with large lesions well-depicted on CT, the corresponding MRI changes were comparable in shape and extent. But MRI gave an additional sign for the presence of liquefactive necrosis, and revealed lesions in the contralateral lobes in two out of three patients with unilateral CT changes. Besides obvious improvement in detection rate, the contributions of MRI to the better selection of treatment protocols, and to the understanding of the pathological process are discussed.

Treatment of nasopharyngeal carcinoma (NPC) is notoriously difficult because of the tumour's highly invasive characteristic and its anatomical proximity to critical structures. To achieve an adequate cancericidal dose to all potential sites of involvement, the dose to the most inferior portions of both temporal lobes will inevitably exceed the usually quoted tolerance limit of 1700 ret (Rottenberg et al., 1977) in terms of nominal standard dose, or 1050 btu (Pezner and Archambeau, 1981) in terms of brain tolerance units. It is not surprising, therefore, that despite the rarity of NPC in Caucasian populations, it is one of the commonest extracranial malignancies leading to late cerebral Tadiation necrosis (Glass et al., 1984). The average incidence of nasopharyngeal carcinoma in Hong Kong is almost 60 times that in Caucasians (Ho, 1982). Due to a serious shortage of radiotherapy machines, schedules consisting of a large dose per fraction were used. This unfortunately further increases the risk of radiation damage (Scheline et al., 1980; Wigg et at., 1981; Pezner and Archambeau, 1981; Cohen and Creditor, 1983). With accumulation of long survivors, the problem of radiation-induced neurological sequelae becomes one of our major concerns, and our series of 102 patients with clinical evidence of late temporal lobe radionecrosis (Lee et al., 1988) becomes one of the largest series reported in the English literature. While surgery is undoubtedly the most definitive way of Correspondence to: Dr Anne W. M. Lee, RadiotherapyDepartment, Queen Elizabeth Hospital, Wylie Road, Kowloon, Hong Kong.

obtaining antemortem histological verification and achieving long-term control, it would be hazardous in our cases because of bilateral involvement. Hence a reliable, non-invasive means of diagnosis and treatment is specially needed. Computed tomography (CT) has been the most helpful investigation (Martins et al., 1977; Mikhael, 1978; Kingsley and Kendall 1981; Shaw and Bates, 1984; Glass et al., 1984; Lee et al., 1988). By correlating the CT changes, clinical features and dose distribution, confident clinical diagnosis can be made in typical cases without resorting to biopsy. Furthermore, our findings show that different CT features reflect different phases in the development of the pathological process. This recognition is important in the selection of treatment, as only patients with marked reactive oedema and minimal liquefactive necrosis are likely to show durable objective response to conservative treatment with corticosteroids (Lee et al., 1988). However, the ability of CT to depict small lesions in the most inferior portions of the temporal lobes is hampered by beam-hardening artefacts from surrounding bone. MRI, with its well-known freedom from such artefacts and exquisite sensitivity for demyelinating diseases (Bydtier et al., 1982; Brant-Zawadzki et al., 1983; Bradley et al., 1984), becomes an attractive alternative. With the recent installation of the first scanner in Hong Kong, we try to assess its relative merits in our cases. This is our preliminary experience in 13 patients.

PATIENTS AND METHODS Patiefit Characteristics

There were eight male and five female patients, with ages ranging from 40 to 63 years (median = 49 years). All had received one radical course of external irradiation for poorly differentiated squamous cell carcinoma of the nasopharynx. None had factors associated with an aggravating effect on radiation damage. All had remained free of any evidence of local recurrence, erosion of the base of the skull and distant metastases. Radiation Dose

All were treated with 4.5 MeV photons using the same technique of two lateral-opposed fields supplemented by an anterior field. The inferomedial parts of both temporal lobes were included within the target volume (Fig. 1). The treatment schedules consisted of 2.5 Gy per fraction, four times weekly to 60 Gy in five patients (Cases 1, 5, 6, 8, 12), and 4.2 Gy twice weekly to 50.4 Gy in the remaining eight.

25

MRI OF TEMPORAL LOBE RADIONECROSIS

(a) Fig. 1 - T h e composite isodose curves, at a plane 20 mm above the principal plane, illustrating the dose distribution to the inferior portions of both temporal lobes.

Thus the estimated doses to the temporal lobes were 1870 and 1861 ret, or 1286 and 1477 btu, respectively. Clinical Picture and CT Changes at Presentation

Two patients (Cases 1, 2) presented with episodes of loss of consciousness, five (Cases 5, 8, 10 to 12) with temporal lobe epilepsy, three (Cases 3, 6, 9) with both, while the remaining three (Cases 4, 7, 13) presented with impairment of memory alone. None complained of severe headache. Neurological examinations revealed no abnormal signs other than nominal aphasia in one patient (Case 7) and varying degrees of amnesia in four (Cases 4, 7, 9, 13). The latent interval from completion of radiotherapy to onset of symptoms ranged from 2 to 11 years (median = 6 years). Cases 1 to 4 were diagnostic problems, as the initial CT failed to reveal any definite abnormalities other than streaking artefacts (Fig. 2a). The remaining nine patients presented with CT changes typical of radiation damage in the inferior portions of the temporal lobes: irregular finger-like hypodense shadows (the finger sign- Fig. 3a) were demonstrated in six patients (Cases 5 to 8, 12, 13) and rounded cyst-like shadows (the cyst sign - Fig, 4a) in three (Cases 9 to 11). The CT changes were bilateral in three patients (Cases 5, 9, 13) and unilateral in the remaining six. Conservative treatment with corticosteroids had been given to all but three patients (Cases 11 to 13). They were regularly reassessed with serial CT and neurological examinations. MRI and CT Methods

MR imaging was performed with a Diasonics tow-field 0.064 T permanent magnet scanner. Sagittal, coronal a n d : axial scans of 5 mm or 10 mm thickness were made, using

(b) Fig. 2 - Case 1. This patient presented with deja vu and episodes of loss o f consciousness. (a) CT fails to show any definite abnormalities other than streaking artefacts. (b) T2-weighted (TR: 2000 ms, TE: 105 ms)

MRI revealsirregularly-shapedlesionsin the most inferiorportions of both temporal lobes.

spin echo sequences of different combinations of repetition time (TR) and echo time (TE), to give Tl-weighted (short TR/short TE), proton-density (tong TR/short TE), and T2-weighted (long TR/long TE) images. CT was performed using a GE 8800 CT/T or GE 9800 CT/T scanner. Axial sections, and in some cases coronal sections were obtained, using a slice thickness of 5 mm or 10 mm. For Cases 5 to 13 with serial CT examinations, the scans done within the same month as the MRI were used for comparison~

26

CLINICAL RADIOLOGY

(a)

(b)

(c) Fig. 3 - Case 5. (a) Initial CT in 1983 shows huge finger signs bilaterally. (b) Follow-up CT in 1988 shows continued complete regression of CT changes 5 years after conservative treatment with corticosteroid. (e) T2-weighted (TR: 2000 ms, TE: 105 ms) MRI reveals small residual abnormalities in the most inferior parts of both temporal lobes.

27

MRI OF TEMPORAL LOBE RADIONECROSIS

(a)

(b)

(c)

(a)

Fig. 4 - Case 9. (a) Axial CT shows huge cyst signs in bilateral temporoparietal regions. (b) Axial T2-weighted spin-echo sequence (TR: 2000 ms, TE: 105 ms) MRI shows very high signal intensity at the same sites. (c) Axial proton-density (TR: 2000 ms, TE: 30 ms) MRI at the same level reveals an area o f low signal intensity within the lesion in the right temporal lobe, highly suggestive of liquefactive changes. (d) Coronal proton-density (TR: 1500 ms, TE: 30 ms) MRI shows the vertical extent of the affected areas.

RESULTS

MRI Findings MRI revealed abnormalities in the white matter in the inferior portions of the temporal lobes in all 13 patients. All manifested as areas of very high signal intensity, best shown on T2-weighted scans (TR of 1000-2000 ms and TE of 105 ms), signifying prolongation of the T2 relaxation time.

All except one patient (Case 11) showed bilateral involvement, but the changes appeared grossly asymmetrical in shape and/or extent in nine of them. The lesions in both temporal lobes appeared irregular in shape in nine patients (Cases 1 to 7, 12, 13), and rounded in one (Case 9). They wdre irregular in one temporal lobe but rounded in the other in the remaining two (Cases 8, 10). In the four patients with rounded lesions (Cases 8 to 11), and in one patient with irregular changes (Case 13), a

28

CLINICAL RADIOLOGY

central area of low signal intensity could be demonstrated within the lesions by a Tl-weighted or proton-density echo sequence (TR of 1000 ms and TE of 30 ms), signifying prolongation of the T1 relaxation time. This feature is highly suggestive of central cavitation with liquefactive changes. Comparison between MRI and CT Changes The results are summarized in Table 1.

Cases 9 to 10

Cases 1 to 4

Bilateral, irregularly-shaped abnormalities welldepicted on MRI (Fig. 2b) were totally overshadowed by artefacts on the corresponding CT (Fig. 2a). MRI was definitely superior to CT in the diagnosis of small radiation lesions confined to the most inferior parts of the temporal lobes. Cases 5 to 7

Their initial finger signs (Fig. 3a) regressed completely after steroid therapy, and they remained free of CT abnormalities for more than 5, 3 and 2 years, respectively. Two of them (Cases 5, 6) still had occasional epileptic attacks despite continuous anticonvulsant treatment. While the most recent CT failed to show any residual or recurrent radiation changes (Fig. 3b), MRI clearly revealed small irregular lesions in the most inferior parts of both temporal lobes (Fig. 3e). The sensitivity of MRI in the detection of small residual lesions after treatment was obvious. Case 8

His presenting symptoms and CT abnormalities completely regressed after conservative treatment with dexamethasone, and he remained in complete remission for 5 Table 1 - Comparison between M R I and C T findings

Cases

Clinical status

CT findings (side/sign)

M R l findings (side/shape)

1 4

Symptomatic new cases - diagnostic problems

None

B irregular

5-6

Steroid treated still symptomatic

No residual finger

Residual B irregular

7

Steroid treated asymptomatic

No residual finger

Residual B irregular

8

Steroid treated asymptomatic

Relapsed L finger + new R cyst

L irregular + R rounded (R necrosis)

9

Steroid treated asymptomatic

Stationary B cyst

B rounded (R necrosis)

10

Steroid treated asymptomatic

Stationary L cyst

L rounded + R irregular (L necrosis)

11

Untreated asymptomatic

Faint residual R cyst

Well-depicted R rounded

12

Untreated symptomatic

Faint residual R finger

Well-depicted B irregular

13

Untreated mildly symptomatic

Stationary B finger

B irregular ( + B necrosis)

B, bilateral. L, left. R, right.

years. Although he was still totally ,asymptomatic, the recent CT revealed reappearance of the initial left-sided finger sign, together with a larger, newly-developed cyst sign on the right. MRI showed similar changes. Information concerning the extent and shape of the lesions was comparable, but MRI gave additional information indicating the presence of central liquefactive necrosis within the rounded lesion.

Both became asymptomatic with treatment by corticosteroids and anticonvulsants, but their huge cyst signs remained unchanged (Fig. 4a) for over 6 and 4 years respectively. These lesions were similarly well depicted by MRI (Fig. 4b), but instead of being apparently homogenous, the proton-density MRI revealed inhomogeneity reflecting different degrees of necrotic changes within the lesions (Figs 4c,d). In addition, while CT only showed unilateral cyst sign in Case D0, MRI revealed a small irregular lesion in the contralateral temporal lobe as well. Cases 11 to 12

Their initial unilateral CT abnormalities (Fig. 5a) showed spontaneous partial regression, leaving very faint hypodense shadows (Fig. 5b) during the subsequent 4 and 2 years of observation respectively. Case 12 still had frequent epileptic attacks. The residual lesions were much better depicted on T2-weighted MRI (Fig. 5e). Furthermore, MRI revealed bilateral involvement in Case 12. Case 13

Her memory problem remained minimal and the small bilateral finger signs on CT remained stationary for over one year without any treatment. T2-weighted MRI also showed bilateral irregularly-shaped lesions of similar extent, but unexpectedly, proton-density MRI revealed small rounded areas within the lesions highly suggestive of central liquefaction (Fig. 6). DISCUSSION There have been only two reports describing the MRI changes of radiation-induced cerebral injury (Dooms et al., 1986; Curran et al., 1987). Both show that MRI is superior to CT in the detection of radiation lesions, which all manifest as areas of high signal intensity on a T2weighted spin echo sequence, signifying prolongation of the T2 relaxation time. However, the changes are nonspecific. As both series were based on patients treated for intracranial neoplasms, and it is impossible to differentiate radiation damages from residual or recurrent tumours from the MRI changes and dose distribution per se, the validity of their conclusions awaits further proof. The present study suppgrts their findings. Although histological verification is also not available in our patients, a confident clinical diagnosis can be made for the following reasons. 1 The bilateral involvement was confined to the inferior parts of the temporal lobes - most unusual for de novo neoplasms, cerebral metastases or abscesses. 2 The absence of co-existing clinical or laboratory findings to indicate an infective element.

MRI OF TEMPORAL LOBE RADIONECROSIS

29

(a)

(b) Fig. 5 - Case 11, (a) Initial CT in 1982 shows a rounded hypodense shadow in the inferior part of the right temporal lobe. (b) Follow-up CT in 1988 shows spontaneous partial reduction. (c, p. 30) T2-weighted (TR: 2000 ms, TE: 105 ms) MRI clearly depicts residual abnormality.

3 The long clinical course and the durable objective response to corticosteroid therapy in particular- most unusual for active tumour or infection. 4 The extreme rarity of haematogenous brain secondaries from nasopharyngeal carcinoma (Ho, 1982). 5 The exclusion o f patients with a possibility of intracranial extension (those with local recurrence or erosion of the base of the skull). The advantage of MRI is especially obvious in our cases as the nidus of radiation damage occurs in an area Where the sensitivity of CT is markedly affected by

artefacts from adjacent bone. M R I revealed definite abnormalities in seven patients with normal CT and two with very faint hypodense shadows; and in two out of three patients with unilateral CT changes, MRI revealed small lesions in the contralateral temporal lobe as well. Indeed, with the exception of one, all patients showed bilateral involvement in the present study with MRI. With this improvement in detection rate, a higher incidence than the previously reported rate of 1.03 % (Lee et al., 1988) is expected. Furthermore, the present findings suggest that while conservative treatment with corticosteroids does have'a retarding influence on the progressive

30

CLINICAL RADIOLOGY

Fig. 5 (c)

Fig. 6 - Case 13. Coronal proton-density (TR: 1000 ms, TE: 30 ms) MRI (left) reveals small rounded areas of low signal intensity within the irregularlyshaped lesions shown on T2-weighted (TR: 1000 ms, TE: 105 ms) MR/(right).

pathological process in patients with marked reactive oedema, genuine complete eradication of the radiation lesions is highly unlikely. In patients with complete disappearance of CT changes," a small nidus of necrosis and/or demyelination may still remain. In patients with large lesions well-depicted on CT, the MRI changes were comparable both in shape and extent. The finger and the cyst signs on CT correspondingly manifested as irregular and rounded lesions on MRI. Hence, it is logical to assume that pathological correla-

tion of the MRI manifestations can be extrapolated from our previous findings based on CT changes (Lee et al., 1988), namely that irregular lesions consist mainly of reactive white matter oedema, and rounded lesions of central liquefactive necrosis with surrounding gliosis. In all four patients with rounded lesions, central areas with prolonged T1 relaxation time, suggesting further increase in water content, could be demonstrated within the lesions by Tl-weighted or proton-density scans. This feature gives another good indicator of central liquefac-

MRI OF TEMPORAL LOBE RADIONECROSIS tion. Its unexpected presence in one p a t i e n t with irregular lesions suggests that some o f the patients with the finger signs on C T m a y h a r b o u r sizeable areas o f necrosis as well as oedema, a n d this p r o b a b l y explains their unresponsiveness to t r e a t m e n t with corticosteroids. In conclusion, M R I is u n d o u b t e d l y superior to C T in sensitivity. N o t only is it i n v a l u a b l e for i m p r o v i n g the detection of late t e m p o r a l lobe d a m a g e in patients with radical r a d i o t h e r a p y for N P C , b u t it also c o n t r i b u t e s significantly to better selection o f t r e a t m e n t protocols, and to the u n d e r s t a n d i n g o f the pathological process. However, it still lacks specificity in differentiating radiation lesions f r o m t u m o u r s a n d in distinguishing between necrosis, d e m y e l i n a t i o n a n d o e d e m a within the lesions. It is hoped that this c a n be i m p r o v e d in future with development o f techniques, using p a r a m a g n e t i c c o n t r a s t agents, spectroscopy or s o d i u m imaging. A t present, C T is still preferable in excluding t u m o u r recurrence with erosion of the base o f the skull. The two investigations should hence be considered c o m p l e m e n t a r y .

Acknowledgements.We thank Ms G. Lau for her secretarialassistance and Mr W. S. Lam for the illustrations.

REFERENCES Bradley, Jr, WG, Waluch, V, Yadley, RA & Wycoff, RR (1984). Comparisonof CT and MR in 400 patients with suspecteddisease of the brain and cervical'spinal cord. Radiology, 152, 695-702. Brant-Zawadzki, M, Davis, PL & Crooks, LE et al. (1983). NMR demonstration of cerebral abnormalities: comparison with CT. American Journal of Radiology, 140, 847-854; American Journal of Neuroradiology, 4, 117-124. • Bydder, GM, Steiner, RE & Young, IR et al. (1982). Clinical NMR imagingof the brain: 140 cases. American Journal of Radiology, 139, 215-236; American Journal of Neuroradiology, 3, 459-480.

31

Cohen, L & Creditor, M (1983). Iso-effect tables for tolerance of irradiated normal human tissues. International Journal of Radiation Oncology, Biology & Physics, 9, 233-241. Curran, WJ, Hecht-Leavitt, C, Schut, L, Zimmerman, RA & Nelson, DF (1987). Magnetic resonance imaging of cranial radiation lesions. International Journal of Radiation Oncology, Biology & Physics, 13, 1093-1098. Dooms, GC, Hecht, S, Brant-Zawadski, M, Berthiaume, Y, Normal, D & Newton, TH (1986). Brain radiation lesions: MR imaging. Radiology, 158, 149-155. Glass, JP, Hwang, TL, Leavens, ME & Libshitz, HI (1984). Cerebral radiation necrosisfollowingtreatment ofextracranial malignancies. Cancer, 54, 1966-1972. Ho, JHC (1982). Treatment of Cancer, pp. 249-267. Chapman & Hall, London. Kingsley, DPE & Kendall, BE (1981). CT of the adverse effects of therapeutic radiation of the central nervous system. American Journal of Neuroradiology, 2, 453-460. Lee, AWM, Ng, SH, Ho, JHC, Tse, VKC, Poon, YF & Tse, CCH et al. (1988). Clinical diagnosis of late temporal lobe necrosis following radiation therapy for nasopharyngeal carcinoma. Cancer, 61, 15351542. Martins, AN, Johnston, JS, Henry, JM, Stoffel, TJ & Di Chiro, G (1977). Delayed radiation necrosis of the brain. Journal of Neurosurgery, 47, 336-345. Mikhael, MA (1978). Radiation necrosis of the brain: correlation between computed tomography, pathology and dose distribution. Journal of Computer Assisted Tomography, 2, 71-80. Pezner, RD & Archambeau, JO (1981). Brain tolerance unit: a method to estimate risk of radiation brain injury for various dose schedules. International Journal of Radiation Oncology, Biology & Physics, 7, 397-402. Rottenberg, DA, Chernik, NL, Deck, MDF, Ellis, F & Posner, JB (1977). Cerebral necrosis following radiotherapy of extracranial neoplasms. Annals of Neurology, 1, 339-357. Scheline, GE, Wara, WM & Smith, V (1980). Therapeutic irradiation and brain injury. International Journal of Radiation Oncology, Biology & Physics, 6, 1215-1228. Shaw, PJ & Bates, D (1984). Conservative treatment of delayedcerebral radiation necrosis. Journal of Neurology, Neurosurgery & Psychiatry, 47, 1338-1341. Wigg, DR, Koschel, K & Hodgson, GS (1981). Tolerance of the mature human central nervous systemto photon irradiation. British Journal of Radiology, 54, 787-798.