- Email: [email protected]
Quantitation of treatment volumes from CT and MRI in high-grade gliomas: Implications for radiotherapy
Mqner;c Printed
Resononce in the USA.
Imagrng, Al1 rights
Vol. 10. pp. 375-383, reserved.
1992 Copyright
0
0730-725X/92 $5.00 + .oO 1992 Pergamon Press Ltd.
??Original Contribution
QUANTITATION HIGH-GRADE
OF TREATMENT VOLUMES FROM CT AND MRI IN GLIOMAS: IMPLICATIONS FOR RADIOTHERAPY
L.C.
MYRIANTHOPOULOS, S. VIJAYAKUMAR, D.R. SPELBRING, S. KRISHNASAMY, S. BLUM, AND G.T.Y. CHEN Michael ReeseNniversity of Chicago Center for Radiation Therapy and Department of Radiation and Cellular Oncology, University of Chicago, Chicago IL 60637, USA
Long-term survival of patients with high-grade gliomas remains extremely poor. The main reason for such an outcome is local failure, or recurrence, after surgery and/or radiotherapy. Higher doses of radiation may result in decreased local failure rates provided that the location (and extent) of gross tumor and microscopic disease can be defined accurately. The abqormalities appearing in images from diagnostic modalities, such as CT and MRI, are being used as a starting point and as a guide for the clinical definition of tumor and its extensions. However, some recent studies on two-dimensional specimens, correlating histopathological findings to CT and MRI images, showed that the resulting definition of tumor cell extensions was unsatisfactory, different, and in need of ample margins. We carried out a retrospective analysis to compare the target volumes that would have been defined by CT, T,-weighted MRI, and Tl-weighted postgadolinium MRI images of the same individual and to explore the implications of the resulting volume definitions for radiotherapy. The results of our limited study, based on the margins used, indicate that the CT-defined target volume is consistently larger than that from either of the two MRI modalities and suggest that noncoplanar approaches for its treatment and other local approaches for tumor boost should be considered. We conclude that until more definitive histopathological guidelines correlated to image features have been formulated and agreed upon, one should try to make full use of all available diagnostic information in order to minimize the possibility of geographical miss of target extensions. Keywords: Three-dimensional treatment planning; Ream’s eye view; Volumetrics; Brain tumor; Glioblastoma multiforme; Radiation therapy; Conformal therapy.
radioresistance of the cancer cells in HGG, tumorrelated factors, such as hypoxia and geographical miss. Most probably failure after RT is due to a combination of al1 these factors. This perhaps is the reason for improved treatment efficacy with larger fields and higher doses. lm6Strategies to improve the local control should address al1 the above factors, but in this paper we wil1 only address possible causes of geographical miss. Historically, HGG used to be treated with limited fields. Data’y4 showing the detrimental effects of field sizes of less than 100 cm2 led to the employment of whole-brain radiotherapy.’ However, with the extensive use of computed tomography (CT) and magnetic resonace imaging (MRI) in the diagnosis and treatment
INTRODUCTION
In spite of recent improvements in the median survival in high-grade gliomas (HGG), long-term survival continues to be dismal. High-grade gliomas in adults rarely metastasize either to the spinal axis or systemically, and they are rarely multicentric. Two HGG characteristics that are of importante in any attempts to improve the outcome of radiotherapy (RT) are the tendency of HGGs to locally infiltrate extensively (vide infra) and their propensity to fail locally after treatment by surgery plus radiotherapy, with or without chemotherapy. Why do HGGs fail locally even after high doses of radiotherapy? The likely causes are the inherent
RECEIVED6/21/91; ACCEPTED10/1/91. Address correspondence to L.C. Myrianthopoulos, Department of Radiation and Cellular Oncology, University
of Chicago Medical Center, 5841 S. Maryland Avenue, Box 442, Chicago IL 60637, USA. 315
316
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
of HGG, there is a tendency to use large-field RT rather than whole-brain RT. This strategy should potentially improve the therapeutic ratio, since less normal brain is being irradiated. However, this can also adversely affect the outcome, if the disease extension defined by CT+ MRI is incorrect. How good are CTand MRI-defined tumor extensions? Daumas-Duport et a1.8 used an improved technique for specimen preparation for histocytopathological examination to study spatial distribution of gliomas from stereotactically obtained biopsy specimens. Three types of spatial distribution were found: (1) Solid tumor (ST) tissue only, (2) centra1 solid tumor surrounded by isolated tumor cells (ITCs), and (3) isolated tumor cells only. The majority of specimens (75%) had the last two patterns: ST + ITC or ITC only. Using this technique, Kelly et a1.9 correlated the histopathological findings to CT-MRI in 39 patients, examining a total of 177 biopsy specimens. About 75% (74/98) of the specimens from hypodense CT areas showed ITC; only 20% (9/45) CT-isodense specimens were normal; 60% (16/23) of Ti-normal and 40% (4/10) of T&normal specimens demonstrated tumor cells. In other words,, the CT-MRI definition of tumor cel1 extensions appears to be poor. Using the same technique, Earnest et al. l” further extended the study to evaluate the usefulness of gadolinium-labeled DTPA regarding its ability to demonstrate the true extent of the disease in MRI studies of gliomas of five untreated patients (four with grade 4 astrocytomas; one with grade 1 mixed glioma) and one previously treated patient. In five out of the six, unenhanced Tz-weighted images had contrast area equal to or more than that of contrast CT; Gd-7’, MRI also showed equivalent or enhanced contrast compared with contrast CT in four out of these six cases. More importantly, biopsy specimens showed solid tumors with neovascularity in three out of the four with grade 4 astrocytoma within CT-MRI contrast enhancement; in two out of two, isolated tumor cells were seen from “sites wel1 beyond” the T2weighted abnormality. CT-contrast regions were “similar but not identical” to Ti-weighted Gd MRI. Contrast CT, contrast Gd MRI, or T,-weighted MRI could not predict tumor boundaries in this limited study of four grade 4 astrocytomas. The limitations of CT in defining the tumor extensions is further confirmed by two papers from Duke University. CT to histopathology correlation was done in whole-brain whole-mount sections in 11 untreated, autopsied patients . l1 Al1 specimens contained one or more areas of central necrosis surrounded by variable width of cellular neoplasm, in turn “surrounded by a halo of neoplasm of lesser cellularity,” that is, ITC.
The outermost area of infiltration varied in extent, and the geometry varied considerably and in three cases extended 3.5, 2.7, and 5.0 cm, respectively. Two-dimensional correlation with CT showed good correlation between CT-contrast enhancement and densely cellular regions. Only 2/12 unilateral lesions were confined within one major cerebral arterial distribution. In 6111, histological tumor was within the CT low-density region, but in 5/11 it extended beyond. The authors concluded that “each Glioblastoma Multiforme must be considered unique in its three-dimensional geometry, particularly in its peripheral portion.” As an extension of the above study, Halperin et al.” planned radiotherapy portals based on CT findings in 11 patients and determined the adequacy of the portals from “topographic distribution of tumor cells” at the time of autopsy. In 9111, the tumor would have been missed geographically if the portals covered the contrast-enhanced area plus a 1 cm margin; in 5/11, if they covered the contrast-enhanced area plus the peritumoral edema with a 1 cm margin; and in none if a 3-cm margin was given. Lesions close to midline commonly crossed the corpus callosum, and tumors spread along nerve pathways. As pointed out by the authors this was a two-dimensional study; and a threedimensional(3D) study would probably show further tumor extensions. The above studies imply the following: (1) it is not yet clear how the microscopic disease (ITC) volume and gross tumor volume (ST) can be defined with CT and/or MRI (*Gd). Further 3D histopathological studies are required for firm recommendations. However, until such studies are completed, we somehow have to define the target volumes in the radiotherapy of HGGs. One approach is shown in Table 1. For the hypothesis in Table 1 (see under Rationale) to be valid, CT- and MRI-defined volumes should be similar. If so, either one of those volumes could be used to plan RT target volumes without any necessity to obtain the second study; if not, then the union of the two volumes defined by the two studies has to be treated to avoid a geographical miss. In the second scenario, both CT and MRI studies need to be performed in al1 patients with the resultant additional tost and time commitments. In addition, image correlation may be necessary. This work was undertaken to compare the target volumes that would have been defined if access to a single diagnostic modality (CT or MRI) was available and to explore the implications of such volume definitions for RT. In particular, the fraction of brain treated and the extent to which geographical miss, in the process of attempting to black the eyes, can be avoided was explicitly considered.
Treatment volumes in high-grade gliomas 0 L.C. MYRLWTHOPOULOS ET
AL.
317
Table 1. Tumor and microscopic disease definitions Volume
Field Initial large field (45-50 Gy)
Boost field (60-70 Gy)
Rationale
CT-visible f peritumoral edema + 3-cm margin (CT LGT)
Studies by Burger et al. and Halperin et al. showing that 3 cm around the peritumoral edema contained cancer cells in al1 patients studied.
Volume of Tl /T2 abnormality + 3-cm margin (MRI LGT)
Study by Kelly et al. showing that the tumor cells were present in 40% of T2 normal specimens; perhaps a margin of 3 cm around it wil1 suffice to make up the uncertainty.
CT-contrast-enhanced volume + 1-cm margin (CT BTM)
Burger et al. showed good correlation between CTcontrast enhancement and densely cellular regions of the tumors.
Ti-weighted Gd-enhanced
Earnest et al. showed CT-contrast and Tr-weighted-
volume
+ 1-cm margin (MRI BTM)
METHODS AND MATERIALS A retrospective analysis was carried out on three patients with an HGG who were referred for radiotherapy at our center. The study was restricted to adults for whom both CT and MRI (with and without Gd contrast) had been performed within a short time interval (a few weeks) of each other prior to initiation of therapy. The digital images were read into our treatment planning system from magnetic tape. For each of three diagnostic modalities, namely contrast-enhanced CT, T,-weighted MRI, and Ti -weighted postgadolinium MRI, tumor and target contours were outlined on successive axial cuts of each patient. Tumor (TUM) was defined as the contrast-enhanced area. Target (TGT) was defined as tumor plus edema and was set equal to tumor if edema was not discernible. In view of the findings of Halperin et al., ‘* a large-field target (LGT) was also entered on each slice with a 3-cm margin around the TGT contour. The LGT outline was entered in along natural barriers such as the inner table of the skull or the falx cerebellum, even if this meant that the margin would be less than 3 cm. Al1 TUM, TGT, and LGT outlines were entered by a radiotherapist following consultation with a neuroradiologist. In addition, the inner table, brain, and eyes were also outlined whenever appropriate. Patients from two different sites of our center were included in this study. Each site employed it own CT and MRI scanners and therefore made use of images of slightly different parameters. However, al1 MRI images were generated using the spin-echo pulse sequence. The T2-weighted images were obtained using TE = 90 msec and TR = 2.2 sec for case A and TE = 70 msec and TR = 2.0 sec for cases B and C. The T,weighted images with gadolinium contrast resulted
contrast regions to be “similar (but not identical)“; see study by Burger et al. above.
from a combination of TE = 19 msec with TR = 0.7 sec for case A and TE = 20 msec and TR = 0.6 sec for cases B and C. The slice thickness of each CT slice was 5.0 mm in the region encompassing each patient’s eyes and was then increased to 10 mm, from about 2 cm cephalad up to the top of the cranium. The CT slice separation was always equal to the corresponding slice thickness, that is, it was also either 5 or 10 mm. The separation and thickness of the T2-weighted images were 9.0 and 6.0 mm, respectively, for case A and 7.5 and 5.0 mm for cases B and C; the values for the í”rweighted images were 10.4 and 8.0 mm for case A and 7.5 and 5.0 mm for cases B and C. The in-plane resolution of the axial MRI images was equal to and usually slightly better than that of the CT images. The pixel size of the latter was constant for a given patient and varied from 0.90 to 0.94 mm in the cases examined. The pixel size of the T2 and Tl images was 0.98 and 0.90 mm for case A and was equal to 0.78 mm for the other cases. Since this was a retrospective study, we had no control over the values of these parameters. However, the computer program that calculated the volumes of interest made proper use of pertinent information. For example, the slice separation, rather than the thickness, was employed in the volume calculations. A typical treatment prescription would be to deliver 45 Gy to the LGT volume and an additional 15 Gy to a boost tumor volume (BTM) defined by TUM plus a 1-cm margin. We determined both the total volumes of the brain and eyes and the variation, as a function of gantry angle, of the fraction of the volume of these organs that would be as irradiated by a field that encompassed either LGT or BTM with a 1-cm margin. We did this using our Beam’s Eye View Volumetrics (BEVOL) computer program. Tests of the accuracy of
378
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
the calculation by BEVOL of the volume of standard geometrical objects, such as a sphere, a tube, and circular and elliptical cylinders, assumed to be intersected by 10 to 12 slices, yielded better than 0.3% agreement between the exact and calculated volumes. The varia-
tion of the fraction of irradiated brain and eye volumes was obtained separately for each of the three diagnostic modalities since we were interested in the effects of differing target and consequently LGT
(4
(Bl
definitions.
Fig. 1. Tumor, target, and large field target outlines for the same patient (case A), drawn over axial cuts. (A) CT image. (B) T,-weighted MRI image. (C) TI-weighted Gd-contrast MRI image. The inner table (brain) is also outlined.
Treatment volumes in high-grade gliomas ??L.C. MYRIANTHOPOULOS ET AL.
RESULTS
319
Table 2. Volumes and fraction range in LGT frame for case A
Case A This is a 71-year-old white woman with diagnosis of glioblastoma multiforme of the left posterofrontal lobe. She underwent partial resection prior to radiotherapy. The CT and MRI studies we used were performed after surgery. Samples of her TUM, TGT, and LGT contours are shown in Fig. 1, for each of the three diagnostic modalities. The fraction of her brain volume that would have been irradiated by a field encompassing each of the three LGTs as a function of BEVOL gantry angle (AP = 90”, RT lateral = 0’) is shown in Fig. 2. It can be seen that a minimum of brain would be treated if approximately 45” opposed right anterior and left posterior oblique fields were to be employed, irrespective of the modality used to define LGT. Moreover, the fraction of brain to be irradiated exceeds 75% (for CT and TJ and 66% (for T,), and, except near the minima, the fraction corresponding to the CT LGT is consistently larger than either of the MRI’s. Figure 3 shows that for the T2 LGT, the angles that spare the most brain would also irradiate a large fraction of both eye volumes and that even lateral opposed (0 and 180”) fields would not
CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
r,
Vol (ml)
Range (%o)
Vol (ml)
Range (Qo)
Vol (ml)
Range W)
1251 58 265 645 5 5
73-100
1285 68 256 624 6 6
75-92
1118 81 427 7 7
66-95
100 0-100 13-100
100 14-100 63-100
completely spare either eye. The absolute volumes and range of volumetric fractions of the various structures and organs of interest for each of the three modalities are presented in Table 2. Case B This is a 57-year-old Chinese man with a diagnosis of glioblastoma multiforme in the left frontoparietal region, extending to the right side through the corpus
12o-l
60 !
100 0-100 21-100
-
Brain CT
-
Brain Gd
-
Brain T2
r
I
90
160
BEVOL “Gantry” Angle
270
(degrees)
Fig. 2. Brain volumetrics for case A for each of the three LGT frames.
360
380
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
J
0 0
-
Lt EyeT2
1
1
90
180
r 270
BEVOL “Gantry” Fig. 3. Brain and eye volumetrics
callosum. The patient refused surgical procedures, including biopsy. In Fig. 4 the brain volumetric variation is compared for the three different modalities. Again, the fractions of brain irradiated are consistently and here significantly larger for the CT frame. Moreover, the minima do not coincide in this case. Figure 5 shows that when framing the T2 LGT frame alone, it would be possible to clear both of this patient’s eyes by treating with bilateral opposed fields (0 and 180’) or with slightly off lateral opposed obliques. Table 3 presents the absolute volumes and fractional ranges for this patient. Case C This is a 40-year-old black woman with grade IIIIV anaplastic astrocytoma of the left media1 frontal lobe with extension to the right frontal lobe through the corpus callosum. In this case, we find consistently larger fractional brain volumes for the CT frame, in phase with the MRI variation. The volumes and ranges pertaining to this patient are presented in Table 4. Table 5 presents the relative ratios of bram, tumor, and LGT volumes defined by the three modalities for each of the three cases considered. The table shows (1)
Angle
360
(degrees)
for case A for the T2 MRI frame.
that brain volumes are acceptably close, (2) that CT and T2-weighted LGT volumes are comparable, and (3) that tumor volume ratios can vary significantly from unity and in either direction. It should be noted that the irradiation of the BTM (as defined in the Methods section) also involves a significant fraction of the brain volume. In case A, for example, framing the CT-defined BTM by 1 cm would treat 34 to 67% of the brain while the Ti-defined
Table 3. Volumes and fraction range in LGT frame for case B CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
G
Vol (ml)
Range
1489 59 134 553 14 13
72-83
(Q)
100 0-11 0-9
Vol (mI1
Range (Qo)
Vol (ml)
Range Po1
1480 83 126 508 9 9
62-67
1482 72 87 443 14 11
63-69
100 0-100 0-100
100 3-100 0-100
Treatment volumes in high-grade gliomas 0 L.C.
0
I
I
90
180
BEVOL “Gantry”
381
ET AL.
Brain CT
-
80:
MYRIANTHOPOULOS
I
270
Angle
:
i0
(degrees)
Fig. 4. Bram volumetrics for case B for each of the three LGT frames.
BTM frame would treat anywhere from 51 to 82% of that patient’s brain. DISCUSSION Identification of problems is the first step toward finding solutions. This study, although limited-, helps to emphasize some of the causes of the dismal results of current treatment of HGGs in adults.
Accurately defining the extent of microscopic disease is stil1 inadequate. In al1 three cases, the microscopic-subclinical disease extent (defined by LGT) differed volumetrically among the studies compared. However, a genera1 trend can be discerned: The CT (LGT) volume is, in al1 three cases, the largest. If contrast CT defines the true extent of microscopic exten-
Table 5. Ratios of volumes normalized to that of CT Table 4. Volumes and fraction range
Modality
in LGT frame for case C
CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
Tz
Case A
Vol (ml)
Range (~01
Vol (ml)
Range (Qo)
Vol (ml)
Range (%1
1177 29 122 528 3 5
82-99
1191 20 107 449 8 8
71-75
1199 73 422 6 6
72-83
B
100 0-60 0-72
C
100 0-42 13-55
100 0-56 0-65
Type of Volume
CT
T, MRI
T, -Gd
Brain Tumor LGT Brain Tumor LGT Brain Tumor LGT
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.03 1.17 0.97 0.99 1.41 0.92 1.01 0.69 0.85
0.89 1.40 0.66 1.00 1.22 0.80 1.02 0.80
382
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
-
Brain T2
-
RT EyeT2
-
Lt FyeT2
0 0
90
180
BEVOL “Gantry” Angle
270
360
(degrees)
Fig. 5. Brain and eye volumetrics for case B for the Tz MRI frame.
sion well, llpl* since the volume of this target (LGT) is
higher than that found when using other modalities, then the fraction of normal brain to be irradiated wil1 be higher and strategies to improve the control of this microscopic disease while sparing normal brain function need to be developed. If CT overestimates the microscopic disease extent, then 3D histopathological studies correlating MRI findings to define the extent of microscopic disease are needed. There is currently an incongruente between the definition of the LGT volumes, as defined by different modalities, and there is a need for further studies to clarify this point. This incongruente is not only limited to volumetrics; a spatial discrepancy can also exist. This is wel1 illustrated when comparing the BEVOL plots for the three cases. Whereas for cases A and C the “phase” is simiIar among the three modalities, in case B the CT volumes are “out-of-phase” with the MRI volumes. The volume of tumor in al1 three cases defined by al1 three modalities is relatively smal1 and constitutes only a smal1 fraction of the brain tissue. Solid tumor, which contains the majority of tumor cel1 burden confined to this volume, needs higher dose for disease control. Brachytherapy is probably suitable for achiev-
ing this goal with the advantage of sparing the normal brain tissue. Recent clinical trials support this concept. However, control of microscopic disease that extensively infiltrates far beyond the confines of the solid tumor, while stil1 sparing the function of normal tissue, needs to be improved. Currently most of the recurrences are infield and occur in the proximity of the tumor bulk. However, as control of this bulky disease improves (e.g., with brachytherapy), unmasking of the “marginal” failure in the area of microscopic disease can be expected for the following reasons: (1) Tumor cells of HGG are more radioresistant than epithelial neoplastic cells defined by conventional in vitro radiosurgical parameters (D,,,N) or parameters defining the radiosensitivity in the clinical dose ranges (surviving fraction at 2 Gy [SF*], mean inactivation dose [i?]).‘3-15 It has been shown that mean values of SF2 categorize various histopathologies from radiosensitive to radioresistant as per clinical experience. 13~14,‘ Dea6 con et al.16 also showed that, based on SF2 values, the cure of epithelial and other relatively radiosensitive (radiocurable) malignancies with doses used in fractionated RT can be explained. It is wel1 known that microscopic-subclinical disease of epithelial origin can
Treatment volumes in high-grade gliomas 0 L.C. MYRUNTHOPOULOS ET AL.
be controlled with 50 Gy. i’***This implies that generally 45-50 Gy applied to microscopic disease in HGG is inadequate for control of the more radioresistant HGG tumor cells. Tumor-related factors (e.g., hypoxia) may further contribute to HGG’s relative radioresistance. l1 From our volumetric analysis, parallel opposed or other coplanar field arrangements are often inadequate for satisfactory geometrie coverage of LGT volume and can easily lead to geometrical miss, especially if one were to irradiate the least fraction of brain to avoid irradiating the eyes in the process. The implication of this is that one needs to consider noncoplanar beam arrangements to improve geometrical coverage and spare normal tissues from irradiation. The need for individualized treatments is also obvious. These considerations point toward the need for both biological (overcoming the inherent radioresistance) and physical (noncoplanar field arrangements, brachytherapy, hyperthermia) approaches to improve the outcome of HGG treatment. From our volumetric study, we conclude the following: (1) Treatment of microscopic disease (as defined by CT-MRI) involves irradiation of over 75% of brain volume, (2) the volume of microscopic disease as defined by CT is consistently greater than that delineated using MRI, with 7”-weighted-MRI-defined microscopic disease coming closest to the CT-defined extent; there is an urgent need to define the significance of CT-MRI findings in 3D, (3) tumor boosts with external beam RT leads to approximately 50% brain volume irradiation; for this reason other local treatment approaches that treat less brain volume, such as brachytherapy and/or hyperthermia, should be pursued vigorously, (4) the constraint of minimizing brain irradiation, while simultaneously avoiding the eyes, may be satisfied better and more easily by noncoplanar approaches, which should therefore be investigated, and (5) until we have more precise histopathological guidelines, correlated to image features in 3D, we should try to combine and use al1 diagnostic information available, in order to avoid possible geo- I graphical miss. Acknowledgment- We thank Dr. Charles Pelizzari, Karen Cross, and James Balter for their multifaceted contributions toward carrying out this work. REFERENCES
1. Arisizibal, S.A.; Caldwell, W.L. Time-dose-volume relationships in the treatment of ghoblastome multiforme. Radiology 101:202-202; 1971. 2. Concannon, J.P.; Kramer, S.; Beny, R. The extent of in-
383
tracranial gliomata at autopsy and its relationship to techniques used in radiation therapy of brain tumors. Am. J. Roentgenol. Radium Ther. Nucl. Med. 84:99107; 1960. 3. Kramer, S. Tumor extent as a determining factor in radiotherapy of glioblastomas. Acts Radiol. (Ther). 8: 111-117; 1969. 4. Salazar, O.M.; Rubin, P., McDonald, J.V., et al.. High dose radiation therapy in the treatment of glioblastoma multiforme: A preliminary report. Int. J. Radiat. Oncol. Bio/. Phys. 1:717-727; 1976. 5. Sheline, G.E. Radiation therapy of primary tumors. Semin. Oncol. 2:29-42; 1975. 6. Todd, I.D.H. Choice of volume in the X-ray treatment of supratentorial gliomas. Br. J. Radiol. 36:645-649; 1963. 7. Kornblith, P.L.; et al. Chapter 9. In: Neurologie Oncology. Philadelphia: J.B. Lippincott; 1987. 8. Daumas-Duport et al., Mayo Clin. Proc. 62:435-449; June 1987. 9. Kelly, et al. Mayo Clin. Proc. 62:450-459; June 1987. 10. Earnest, F., IV; Kelly, P.J.; Scheithauer, B.W.; Kali, B.A.; Cascino, T.L.; Ehman, R.L.; Forbes, G.S.; Axley, P.L. Cerebral astrocytomas: Histopathologie correlation of MR and CT contrast enhancement with stereotactic biopsy. Radiology 166:823-827; 1988. 11. Burger, P.C.; Heinz, E.R.; Shibata, T.; Kleihues, P. Topographic anatomy and CT correlations in the untreated glioblastoma multiforme. J. Neurosurg. 68:698704; 1988. 12. Halperin, E.C.; Bentel, G. ; Heinz, E.R.; Burger, P.C. Radiation therapy treatment planning in supratentorial glioblastoma multiforme: An analysis based on post mortem topographic anatomy with CT correlations. Int. J. Radiat. Oncol. Biol. Phys. 17:1347-1350; 1989. 13. Fertil, B.; Malaise, E.P. Inherent cellular radiosensitivity as a basic concept for human tumor radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 7:621-629; 1981. 14. Fertil, B.; Malaise, E.P. Intrinsic radiosensitivity of human cel1 lines is correlated with radioresponsiveness of human tumors: Analysis of 101 published survival curves. Int. J. Radiat. Oncol. Biol. Phys. 11: 1699-1707; 1985. 15. Vijayakumar, S.; Ng, T.C.; Raudkivi, U.; Meaney, T. J. Mean inactivation dose (D): A critical analysis of a neglected parameter in radiotherapy. Acts Oncol. 29:6572; 1990. 16. Deacon, J.; Peckham, M.J.; Steel, G.G. The radioresponsiveness of human tumours and the initial slope of the cel1 survival curve. Radiother. Oncol. 2:317-323; 1984. 17. Fletcher, G.H. The scientific basis of the present and future practice of clinical radiotheraphy. Int. J. Radiat. Oncol. Biel. Phys. 9:1073-1082; 1983. 18. Fletcher, G.H. Implications of the density of clonogenic infestation in radiotherapy. Int. J. Radiat. Oncol. Bio/. Phys. 12:1675-1680; 1986.
Resononce in the USA.
Imagrng, Al1 rights
Vol. 10. pp. 375-383, reserved.
1992 Copyright
0
0730-725X/92 $5.00 + .oO 1992 Pergamon Press Ltd.
??Original Contribution
QUANTITATION HIGH-GRADE
OF TREATMENT VOLUMES FROM CT AND MRI IN GLIOMAS: IMPLICATIONS FOR RADIOTHERAPY
L.C.
MYRIANTHOPOULOS, S. VIJAYAKUMAR, D.R. SPELBRING, S. KRISHNASAMY, S. BLUM, AND G.T.Y. CHEN Michael ReeseNniversity of Chicago Center for Radiation Therapy and Department of Radiation and Cellular Oncology, University of Chicago, Chicago IL 60637, USA
Long-term survival of patients with high-grade gliomas remains extremely poor. The main reason for such an outcome is local failure, or recurrence, after surgery and/or radiotherapy. Higher doses of radiation may result in decreased local failure rates provided that the location (and extent) of gross tumor and microscopic disease can be defined accurately. The abqormalities appearing in images from diagnostic modalities, such as CT and MRI, are being used as a starting point and as a guide for the clinical definition of tumor and its extensions. However, some recent studies on two-dimensional specimens, correlating histopathological findings to CT and MRI images, showed that the resulting definition of tumor cell extensions was unsatisfactory, different, and in need of ample margins. We carried out a retrospective analysis to compare the target volumes that would have been defined by CT, T,-weighted MRI, and Tl-weighted postgadolinium MRI images of the same individual and to explore the implications of the resulting volume definitions for radiotherapy. The results of our limited study, based on the margins used, indicate that the CT-defined target volume is consistently larger than that from either of the two MRI modalities and suggest that noncoplanar approaches for its treatment and other local approaches for tumor boost should be considered. We conclude that until more definitive histopathological guidelines correlated to image features have been formulated and agreed upon, one should try to make full use of all available diagnostic information in order to minimize the possibility of geographical miss of target extensions. Keywords: Three-dimensional treatment planning; Ream’s eye view; Volumetrics; Brain tumor; Glioblastoma multiforme; Radiation therapy; Conformal therapy.
radioresistance of the cancer cells in HGG, tumorrelated factors, such as hypoxia and geographical miss. Most probably failure after RT is due to a combination of al1 these factors. This perhaps is the reason for improved treatment efficacy with larger fields and higher doses. lm6Strategies to improve the local control should address al1 the above factors, but in this paper we wil1 only address possible causes of geographical miss. Historically, HGG used to be treated with limited fields. Data’y4 showing the detrimental effects of field sizes of less than 100 cm2 led to the employment of whole-brain radiotherapy.’ However, with the extensive use of computed tomography (CT) and magnetic resonace imaging (MRI) in the diagnosis and treatment
INTRODUCTION
In spite of recent improvements in the median survival in high-grade gliomas (HGG), long-term survival continues to be dismal. High-grade gliomas in adults rarely metastasize either to the spinal axis or systemically, and they are rarely multicentric. Two HGG characteristics that are of importante in any attempts to improve the outcome of radiotherapy (RT) are the tendency of HGGs to locally infiltrate extensively (vide infra) and their propensity to fail locally after treatment by surgery plus radiotherapy, with or without chemotherapy. Why do HGGs fail locally even after high doses of radiotherapy? The likely causes are the inherent
RECEIVED6/21/91; ACCEPTED10/1/91. Address correspondence to L.C. Myrianthopoulos, Department of Radiation and Cellular Oncology, University
of Chicago Medical Center, 5841 S. Maryland Avenue, Box 442, Chicago IL 60637, USA. 315
316
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
of HGG, there is a tendency to use large-field RT rather than whole-brain RT. This strategy should potentially improve the therapeutic ratio, since less normal brain is being irradiated. However, this can also adversely affect the outcome, if the disease extension defined by CT+ MRI is incorrect. How good are CTand MRI-defined tumor extensions? Daumas-Duport et a1.8 used an improved technique for specimen preparation for histocytopathological examination to study spatial distribution of gliomas from stereotactically obtained biopsy specimens. Three types of spatial distribution were found: (1) Solid tumor (ST) tissue only, (2) centra1 solid tumor surrounded by isolated tumor cells (ITCs), and (3) isolated tumor cells only. The majority of specimens (75%) had the last two patterns: ST + ITC or ITC only. Using this technique, Kelly et a1.9 correlated the histopathological findings to CT-MRI in 39 patients, examining a total of 177 biopsy specimens. About 75% (74/98) of the specimens from hypodense CT areas showed ITC; only 20% (9/45) CT-isodense specimens were normal; 60% (16/23) of Ti-normal and 40% (4/10) of T&normal specimens demonstrated tumor cells. In other words,, the CT-MRI definition of tumor cel1 extensions appears to be poor. Using the same technique, Earnest et al. l” further extended the study to evaluate the usefulness of gadolinium-labeled DTPA regarding its ability to demonstrate the true extent of the disease in MRI studies of gliomas of five untreated patients (four with grade 4 astrocytomas; one with grade 1 mixed glioma) and one previously treated patient. In five out of the six, unenhanced Tz-weighted images had contrast area equal to or more than that of contrast CT; Gd-7’, MRI also showed equivalent or enhanced contrast compared with contrast CT in four out of these six cases. More importantly, biopsy specimens showed solid tumors with neovascularity in three out of the four with grade 4 astrocytoma within CT-MRI contrast enhancement; in two out of two, isolated tumor cells were seen from “sites wel1 beyond” the T2weighted abnormality. CT-contrast regions were “similar but not identical” to Ti-weighted Gd MRI. Contrast CT, contrast Gd MRI, or T,-weighted MRI could not predict tumor boundaries in this limited study of four grade 4 astrocytomas. The limitations of CT in defining the tumor extensions is further confirmed by two papers from Duke University. CT to histopathology correlation was done in whole-brain whole-mount sections in 11 untreated, autopsied patients . l1 Al1 specimens contained one or more areas of central necrosis surrounded by variable width of cellular neoplasm, in turn “surrounded by a halo of neoplasm of lesser cellularity,” that is, ITC.
The outermost area of infiltration varied in extent, and the geometry varied considerably and in three cases extended 3.5, 2.7, and 5.0 cm, respectively. Two-dimensional correlation with CT showed good correlation between CT-contrast enhancement and densely cellular regions. Only 2/12 unilateral lesions were confined within one major cerebral arterial distribution. In 6111, histological tumor was within the CT low-density region, but in 5/11 it extended beyond. The authors concluded that “each Glioblastoma Multiforme must be considered unique in its three-dimensional geometry, particularly in its peripheral portion.” As an extension of the above study, Halperin et al.” planned radiotherapy portals based on CT findings in 11 patients and determined the adequacy of the portals from “topographic distribution of tumor cells” at the time of autopsy. In 9111, the tumor would have been missed geographically if the portals covered the contrast-enhanced area plus a 1 cm margin; in 5/11, if they covered the contrast-enhanced area plus the peritumoral edema with a 1 cm margin; and in none if a 3-cm margin was given. Lesions close to midline commonly crossed the corpus callosum, and tumors spread along nerve pathways. As pointed out by the authors this was a two-dimensional study; and a threedimensional(3D) study would probably show further tumor extensions. The above studies imply the following: (1) it is not yet clear how the microscopic disease (ITC) volume and gross tumor volume (ST) can be defined with CT and/or MRI (*Gd). Further 3D histopathological studies are required for firm recommendations. However, until such studies are completed, we somehow have to define the target volumes in the radiotherapy of HGGs. One approach is shown in Table 1. For the hypothesis in Table 1 (see under Rationale) to be valid, CT- and MRI-defined volumes should be similar. If so, either one of those volumes could be used to plan RT target volumes without any necessity to obtain the second study; if not, then the union of the two volumes defined by the two studies has to be treated to avoid a geographical miss. In the second scenario, both CT and MRI studies need to be performed in al1 patients with the resultant additional tost and time commitments. In addition, image correlation may be necessary. This work was undertaken to compare the target volumes that would have been defined if access to a single diagnostic modality (CT or MRI) was available and to explore the implications of such volume definitions for RT. In particular, the fraction of brain treated and the extent to which geographical miss, in the process of attempting to black the eyes, can be avoided was explicitly considered.
Treatment volumes in high-grade gliomas 0 L.C. MYRLWTHOPOULOS ET
AL.
317
Table 1. Tumor and microscopic disease definitions Volume
Field Initial large field (45-50 Gy)
Boost field (60-70 Gy)
Rationale
CT-visible f peritumoral edema + 3-cm margin (CT LGT)
Studies by Burger et al. and Halperin et al. showing that 3 cm around the peritumoral edema contained cancer cells in al1 patients studied.
Volume of Tl /T2 abnormality + 3-cm margin (MRI LGT)
Study by Kelly et al. showing that the tumor cells were present in 40% of T2 normal specimens; perhaps a margin of 3 cm around it wil1 suffice to make up the uncertainty.
CT-contrast-enhanced volume + 1-cm margin (CT BTM)
Burger et al. showed good correlation between CTcontrast enhancement and densely cellular regions of the tumors.
Ti-weighted Gd-enhanced
Earnest et al. showed CT-contrast and Tr-weighted-
volume
+ 1-cm margin (MRI BTM)
METHODS AND MATERIALS A retrospective analysis was carried out on three patients with an HGG who were referred for radiotherapy at our center. The study was restricted to adults for whom both CT and MRI (with and without Gd contrast) had been performed within a short time interval (a few weeks) of each other prior to initiation of therapy. The digital images were read into our treatment planning system from magnetic tape. For each of three diagnostic modalities, namely contrast-enhanced CT, T,-weighted MRI, and Ti -weighted postgadolinium MRI, tumor and target contours were outlined on successive axial cuts of each patient. Tumor (TUM) was defined as the contrast-enhanced area. Target (TGT) was defined as tumor plus edema and was set equal to tumor if edema was not discernible. In view of the findings of Halperin et al., ‘* a large-field target (LGT) was also entered on each slice with a 3-cm margin around the TGT contour. The LGT outline was entered in along natural barriers such as the inner table of the skull or the falx cerebellum, even if this meant that the margin would be less than 3 cm. Al1 TUM, TGT, and LGT outlines were entered by a radiotherapist following consultation with a neuroradiologist. In addition, the inner table, brain, and eyes were also outlined whenever appropriate. Patients from two different sites of our center were included in this study. Each site employed it own CT and MRI scanners and therefore made use of images of slightly different parameters. However, al1 MRI images were generated using the spin-echo pulse sequence. The T2-weighted images were obtained using TE = 90 msec and TR = 2.2 sec for case A and TE = 70 msec and TR = 2.0 sec for cases B and C. The T,weighted images with gadolinium contrast resulted
contrast regions to be “similar (but not identical)“; see study by Burger et al. above.
from a combination of TE = 19 msec with TR = 0.7 sec for case A and TE = 20 msec and TR = 0.6 sec for cases B and C. The slice thickness of each CT slice was 5.0 mm in the region encompassing each patient’s eyes and was then increased to 10 mm, from about 2 cm cephalad up to the top of the cranium. The CT slice separation was always equal to the corresponding slice thickness, that is, it was also either 5 or 10 mm. The separation and thickness of the T2-weighted images were 9.0 and 6.0 mm, respectively, for case A and 7.5 and 5.0 mm for cases B and C; the values for the í”rweighted images were 10.4 and 8.0 mm for case A and 7.5 and 5.0 mm for cases B and C. The in-plane resolution of the axial MRI images was equal to and usually slightly better than that of the CT images. The pixel size of the latter was constant for a given patient and varied from 0.90 to 0.94 mm in the cases examined. The pixel size of the T2 and Tl images was 0.98 and 0.90 mm for case A and was equal to 0.78 mm for the other cases. Since this was a retrospective study, we had no control over the values of these parameters. However, the computer program that calculated the volumes of interest made proper use of pertinent information. For example, the slice separation, rather than the thickness, was employed in the volume calculations. A typical treatment prescription would be to deliver 45 Gy to the LGT volume and an additional 15 Gy to a boost tumor volume (BTM) defined by TUM plus a 1-cm margin. We determined both the total volumes of the brain and eyes and the variation, as a function of gantry angle, of the fraction of the volume of these organs that would be as irradiated by a field that encompassed either LGT or BTM with a 1-cm margin. We did this using our Beam’s Eye View Volumetrics (BEVOL) computer program. Tests of the accuracy of
378
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
the calculation by BEVOL of the volume of standard geometrical objects, such as a sphere, a tube, and circular and elliptical cylinders, assumed to be intersected by 10 to 12 slices, yielded better than 0.3% agreement between the exact and calculated volumes. The varia-
tion of the fraction of irradiated brain and eye volumes was obtained separately for each of the three diagnostic modalities since we were interested in the effects of differing target and consequently LGT
(4
(Bl
definitions.
Fig. 1. Tumor, target, and large field target outlines for the same patient (case A), drawn over axial cuts. (A) CT image. (B) T,-weighted MRI image. (C) TI-weighted Gd-contrast MRI image. The inner table (brain) is also outlined.
Treatment volumes in high-grade gliomas ??L.C. MYRIANTHOPOULOS ET AL.
RESULTS
319
Table 2. Volumes and fraction range in LGT frame for case A
Case A This is a 71-year-old white woman with diagnosis of glioblastoma multiforme of the left posterofrontal lobe. She underwent partial resection prior to radiotherapy. The CT and MRI studies we used were performed after surgery. Samples of her TUM, TGT, and LGT contours are shown in Fig. 1, for each of the three diagnostic modalities. The fraction of her brain volume that would have been irradiated by a field encompassing each of the three LGTs as a function of BEVOL gantry angle (AP = 90”, RT lateral = 0’) is shown in Fig. 2. It can be seen that a minimum of brain would be treated if approximately 45” opposed right anterior and left posterior oblique fields were to be employed, irrespective of the modality used to define LGT. Moreover, the fraction of brain to be irradiated exceeds 75% (for CT and TJ and 66% (for T,), and, except near the minima, the fraction corresponding to the CT LGT is consistently larger than either of the MRI’s. Figure 3 shows that for the T2 LGT, the angles that spare the most brain would also irradiate a large fraction of both eye volumes and that even lateral opposed (0 and 180”) fields would not
CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
r,
Vol (ml)
Range (%o)
Vol (ml)
Range (Qo)
Vol (ml)
Range W)
1251 58 265 645 5 5
73-100
1285 68 256 624 6 6
75-92
1118 81 427 7 7
66-95
100 0-100 13-100
100 14-100 63-100
completely spare either eye. The absolute volumes and range of volumetric fractions of the various structures and organs of interest for each of the three modalities are presented in Table 2. Case B This is a 57-year-old Chinese man with a diagnosis of glioblastoma multiforme in the left frontoparietal region, extending to the right side through the corpus
12o-l
60 !
100 0-100 21-100
-
Brain CT
-
Brain Gd
-
Brain T2
r
I
90
160
BEVOL “Gantry” Angle
270
(degrees)
Fig. 2. Brain volumetrics for case A for each of the three LGT frames.
360
380
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
J
0 0
-
Lt EyeT2
1
1
90
180
r 270
BEVOL “Gantry” Fig. 3. Brain and eye volumetrics
callosum. The patient refused surgical procedures, including biopsy. In Fig. 4 the brain volumetric variation is compared for the three different modalities. Again, the fractions of brain irradiated are consistently and here significantly larger for the CT frame. Moreover, the minima do not coincide in this case. Figure 5 shows that when framing the T2 LGT frame alone, it would be possible to clear both of this patient’s eyes by treating with bilateral opposed fields (0 and 180’) or with slightly off lateral opposed obliques. Table 3 presents the absolute volumes and fractional ranges for this patient. Case C This is a 40-year-old black woman with grade IIIIV anaplastic astrocytoma of the left media1 frontal lobe with extension to the right frontal lobe through the corpus callosum. In this case, we find consistently larger fractional brain volumes for the CT frame, in phase with the MRI variation. The volumes and ranges pertaining to this patient are presented in Table 4. Table 5 presents the relative ratios of bram, tumor, and LGT volumes defined by the three modalities for each of the three cases considered. The table shows (1)
Angle
360
(degrees)
for case A for the T2 MRI frame.
that brain volumes are acceptably close, (2) that CT and T2-weighted LGT volumes are comparable, and (3) that tumor volume ratios can vary significantly from unity and in either direction. It should be noted that the irradiation of the BTM (as defined in the Methods section) also involves a significant fraction of the brain volume. In case A, for example, framing the CT-defined BTM by 1 cm would treat 34 to 67% of the brain while the Ti-defined
Table 3. Volumes and fraction range in LGT frame for case B CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
G
Vol (ml)
Range
1489 59 134 553 14 13
72-83
(Q)
100 0-11 0-9
Vol (mI1
Range (Qo)
Vol (ml)
Range Po1
1480 83 126 508 9 9
62-67
1482 72 87 443 14 11
63-69
100 0-100 0-100
100 3-100 0-100
Treatment volumes in high-grade gliomas 0 L.C.
0
I
I
90
180
BEVOL “Gantry”
381
ET AL.
Brain CT
-
80:
MYRIANTHOPOULOS
I
270
Angle
:
i0
(degrees)
Fig. 4. Bram volumetrics for case B for each of the three LGT frames.
BTM frame would treat anywhere from 51 to 82% of that patient’s brain. DISCUSSION Identification of problems is the first step toward finding solutions. This study, although limited-, helps to emphasize some of the causes of the dismal results of current treatment of HGGs in adults.
Accurately defining the extent of microscopic disease is stil1 inadequate. In al1 three cases, the microscopic-subclinical disease extent (defined by LGT) differed volumetrically among the studies compared. However, a genera1 trend can be discerned: The CT (LGT) volume is, in al1 three cases, the largest. If contrast CT defines the true extent of microscopic exten-
Table 5. Ratios of volumes normalized to that of CT Table 4. Volumes and fraction range
Modality
in LGT frame for case C
CT
Brain Tumor Target LGT Rt Eye Lt Eye
T, -Gd
Tz
Case A
Vol (ml)
Range (~01
Vol (ml)
Range (Qo)
Vol (ml)
Range (%1
1177 29 122 528 3 5
82-99
1191 20 107 449 8 8
71-75
1199 73 422 6 6
72-83
B
100 0-60 0-72
C
100 0-42 13-55
100 0-56 0-65
Type of Volume
CT
T, MRI
T, -Gd
Brain Tumor LGT Brain Tumor LGT Brain Tumor LGT
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.03 1.17 0.97 0.99 1.41 0.92 1.01 0.69 0.85
0.89 1.40 0.66 1.00 1.22 0.80 1.02 0.80
382
Magnetic Resonance Imaging 0 Volume 10, Number 3, 1992
-
Brain T2
-
RT EyeT2
-
Lt FyeT2
0 0
90
180
BEVOL “Gantry” Angle
270
360
(degrees)
Fig. 5. Brain and eye volumetrics for case B for the Tz MRI frame.
sion well, llpl* since the volume of this target (LGT) is
higher than that found when using other modalities, then the fraction of normal brain to be irradiated wil1 be higher and strategies to improve the control of this microscopic disease while sparing normal brain function need to be developed. If CT overestimates the microscopic disease extent, then 3D histopathological studies correlating MRI findings to define the extent of microscopic disease are needed. There is currently an incongruente between the definition of the LGT volumes, as defined by different modalities, and there is a need for further studies to clarify this point. This incongruente is not only limited to volumetrics; a spatial discrepancy can also exist. This is wel1 illustrated when comparing the BEVOL plots for the three cases. Whereas for cases A and C the “phase” is simiIar among the three modalities, in case B the CT volumes are “out-of-phase” with the MRI volumes. The volume of tumor in al1 three cases defined by al1 three modalities is relatively smal1 and constitutes only a smal1 fraction of the brain tissue. Solid tumor, which contains the majority of tumor cel1 burden confined to this volume, needs higher dose for disease control. Brachytherapy is probably suitable for achiev-
ing this goal with the advantage of sparing the normal brain tissue. Recent clinical trials support this concept. However, control of microscopic disease that extensively infiltrates far beyond the confines of the solid tumor, while stil1 sparing the function of normal tissue, needs to be improved. Currently most of the recurrences are infield and occur in the proximity of the tumor bulk. However, as control of this bulky disease improves (e.g., with brachytherapy), unmasking of the “marginal” failure in the area of microscopic disease can be expected for the following reasons: (1) Tumor cells of HGG are more radioresistant than epithelial neoplastic cells defined by conventional in vitro radiosurgical parameters (D,,,N) or parameters defining the radiosensitivity in the clinical dose ranges (surviving fraction at 2 Gy [SF*], mean inactivation dose [i?]).‘3-15 It has been shown that mean values of SF2 categorize various histopathologies from radiosensitive to radioresistant as per clinical experience. 13~14,‘ Dea6 con et al.16 also showed that, based on SF2 values, the cure of epithelial and other relatively radiosensitive (radiocurable) malignancies with doses used in fractionated RT can be explained. It is wel1 known that microscopic-subclinical disease of epithelial origin can
Treatment volumes in high-grade gliomas 0 L.C. MYRUNTHOPOULOS ET AL.
be controlled with 50 Gy. i’***This implies that generally 45-50 Gy applied to microscopic disease in HGG is inadequate for control of the more radioresistant HGG tumor cells. Tumor-related factors (e.g., hypoxia) may further contribute to HGG’s relative radioresistance. l1 From our volumetric analysis, parallel opposed or other coplanar field arrangements are often inadequate for satisfactory geometrie coverage of LGT volume and can easily lead to geometrical miss, especially if one were to irradiate the least fraction of brain to avoid irradiating the eyes in the process. The implication of this is that one needs to consider noncoplanar beam arrangements to improve geometrical coverage and spare normal tissues from irradiation. The need for individualized treatments is also obvious. These considerations point toward the need for both biological (overcoming the inherent radioresistance) and physical (noncoplanar field arrangements, brachytherapy, hyperthermia) approaches to improve the outcome of HGG treatment. From our volumetric study, we conclude the following: (1) Treatment of microscopic disease (as defined by CT-MRI) involves irradiation of over 75% of brain volume, (2) the volume of microscopic disease as defined by CT is consistently greater than that delineated using MRI, with 7”-weighted-MRI-defined microscopic disease coming closest to the CT-defined extent; there is an urgent need to define the significance of CT-MRI findings in 3D, (3) tumor boosts with external beam RT leads to approximately 50% brain volume irradiation; for this reason other local treatment approaches that treat less brain volume, such as brachytherapy and/or hyperthermia, should be pursued vigorously, (4) the constraint of minimizing brain irradiation, while simultaneously avoiding the eyes, may be satisfied better and more easily by noncoplanar approaches, which should therefore be investigated, and (5) until we have more precise histopathological guidelines, correlated to image features in 3D, we should try to combine and use al1 diagnostic information available, in order to avoid possible geo- I graphical miss. Acknowledgment- We thank Dr. Charles Pelizzari, Karen Cross, and James Balter for their multifaceted contributions toward carrying out this work. REFERENCES
1. Arisizibal, S.A.; Caldwell, W.L. Time-dose-volume relationships in the treatment of ghoblastome multiforme. Radiology 101:202-202; 1971. 2. Concannon, J.P.; Kramer, S.; Beny, R. The extent of in-
383
tracranial gliomata at autopsy and its relationship to techniques used in radiation therapy of brain tumors. Am. J. Roentgenol. Radium Ther. Nucl. Med. 84:99107; 1960. 3. Kramer, S. Tumor extent as a determining factor in radiotherapy of glioblastomas. Acts Radiol. (Ther). 8: 111-117; 1969. 4. Salazar, O.M.; Rubin, P., McDonald, J.V., et al.. High dose radiation therapy in the treatment of glioblastoma multiforme: A preliminary report. Int. J. Radiat. Oncol. Bio/. Phys. 1:717-727; 1976. 5. Sheline, G.E. Radiation therapy of primary tumors. Semin. Oncol. 2:29-42; 1975. 6. Todd, I.D.H. Choice of volume in the X-ray treatment of supratentorial gliomas. Br. J. Radiol. 36:645-649; 1963. 7. Kornblith, P.L.; et al. Chapter 9. In: Neurologie Oncology. Philadelphia: J.B. Lippincott; 1987. 8. Daumas-Duport et al., Mayo Clin. Proc. 62:435-449; June 1987. 9. Kelly, et al. Mayo Clin. Proc. 62:450-459; June 1987. 10. Earnest, F., IV; Kelly, P.J.; Scheithauer, B.W.; Kali, B.A.; Cascino, T.L.; Ehman, R.L.; Forbes, G.S.; Axley, P.L. Cerebral astrocytomas: Histopathologie correlation of MR and CT contrast enhancement with stereotactic biopsy. Radiology 166:823-827; 1988. 11. Burger, P.C.; Heinz, E.R.; Shibata, T.; Kleihues, P. Topographic anatomy and CT correlations in the untreated glioblastoma multiforme. J. Neurosurg. 68:698704; 1988. 12. Halperin, E.C.; Bentel, G. ; Heinz, E.R.; Burger, P.C. Radiation therapy treatment planning in supratentorial glioblastoma multiforme: An analysis based on post mortem topographic anatomy with CT correlations. Int. J. Radiat. Oncol. Biol. Phys. 17:1347-1350; 1989. 13. Fertil, B.; Malaise, E.P. Inherent cellular radiosensitivity as a basic concept for human tumor radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 7:621-629; 1981. 14. Fertil, B.; Malaise, E.P. Intrinsic radiosensitivity of human cel1 lines is correlated with radioresponsiveness of human tumors: Analysis of 101 published survival curves. Int. J. Radiat. Oncol. Biol. Phys. 11: 1699-1707; 1985. 15. Vijayakumar, S.; Ng, T.C.; Raudkivi, U.; Meaney, T. J. Mean inactivation dose (D): A critical analysis of a neglected parameter in radiotherapy. Acts Oncol. 29:6572; 1990. 16. Deacon, J.; Peckham, M.J.; Steel, G.G. The radioresponsiveness of human tumours and the initial slope of the cel1 survival curve. Radiother. Oncol. 2:317-323; 1984. 17. Fletcher, G.H. The scientific basis of the present and future practice of clinical radiotheraphy. Int. J. Radiat. Oncol. Biel. Phys. 9:1073-1082; 1983. 18. Fletcher, G.H. Implications of the density of clonogenic infestation in radiotherapy. Int. J. Radiat. Oncol. Bio/. Phys. 12:1675-1680; 1986.