Evaluating Pediatric Brain Tumor Cellularity with Diffusion-Tensor Imaging
Abstract
OBJECTIVE. MR imaging of central nervous system (CNS) malignancies falls short of a definitive evaluation. Tissue diagnosis remains the gold standard. Diffusion-tensor MR imaging measures the apparent diffusion coefficient and diffusion anisotropy of water in tissue. The purpose of this study was to test the hypothesis that the apparent diffusion coefficient may improve the MR imaging evaluation of newly diagnosed CNS neoplasms. We examined the relationship between the apparent diffusion coefficient, anisotropy, and tumor cellularity in 12 pediatric patients.
MATERIALS AND METHODS. On the basis of histopathologic evaluation, tumors in this case series were segregated into three types: low-grade gliomas, embryonal tumors, and nonembryonal high-grade tumors. Mean apparent diffusion coefficient and anisotropy values obtained from the solid components of each tumor were compared with cellularity, total cellular area, and total nuclear area derived from biopsy material.
RESULTS. The apparent diffusion coefficient ratio (tumor to normal brain) correlated well with tumor classification (p = 0.001). Anisotropy was decreased similarly in all tumor classifications. The absolute apparent diffusion coefficient correlated well with cellularity (p = 0.014) and total nuclear area (p = 0.005) per high-power field. The correlation between apparent diffusion coefficient and total cellular area per high-power field was not statistically significant.
CONCLUSION. The apparent diffusion coefficient may be predictive of tumor classification and may be a useful tool in characterizing tumor cellularity and total nuclear area. These parameters are not available in standard MR imaging. Therefore, diffusion-tensor imaging may enhance the diagnostic process in pediatric CNS malignancies.
Introduction
Pediatric central nervous system (CNS) tumors encompass a heterogeneous group of cell types [1]. Clinical treatment and prognosis varies with tumor type, grade, stage, and extent of resection. Conventional MR imaging incorporating contrast-enhanced T1-weighted and T2-weighted sequences helps to characterize the location and extent of these tumors, but MR imaging provides limited information regarding tumor type and grade. Consequently, conventional MR imaging falls short as a definitive diagnostic examination; that role is reserved for histopathologic evaluation after biopsy.
Diffusion imaging produces images in which relative intensity is related to microscopic displacement of water molecules [2]. Diffusion-tensor imaging is a more sophisticated quantitative form of diffusion imaging. For diffusion-tensor imaging, water diffusion is measured in several orientations relative to the sample. When these measurements are combined, it is possible to calculate not only an absolute measure of average water diffusion for each voxel (the apparent diffusion coefficient) but also a measure of how water diffusion varies along different axes of the image (diffusion anisotropy). Water diffusion is anisotropic in white matter, for example, with diffusion coefficients measured parallel to nerve fibers being higher than those measured perpendicular to them. Overall, both apparent diffusion coefficient and diffusion anisotropy reflect the microstructure of the tissue in which they are measured.
The purpose of this study was to determine if quantitative diffusion-tensor imaging improves the MR imaging of pediatric CNS malignancies in vivo by comparing the tumor apparent diffusion coefficient to light microscopic features. We hypothesized that the most cellular tumors and those with the least amount of cytoplasm (greater nuclear area) would have the lowest apparent diffusion coefficient values. This information could play an important adjunctive role in the workup and treatment of pediatric CNS neoplasms.
Materials and Methods
The protocol described in our study was approved by the institutional review board at our institution. As a first step, diffusion-tensor imaging was added to the tumor MR imaging protocol. We then retrospectively reviewed our clinical experience over an 18-month period for tumors that met predefined inclusion criteria. Each pediatric patient enrolled in our study had preresection pretreatment MR imaging at our institution that included diffusion-tensor imaging. Furthermore, each patient included in the study had an intraaxial mass with solid portions suitable for region of interest (ROI) analysis. Finally, the resection should have been performed at our institution to enable histopathologic analysis of the specimen. Patients who had preoperative examinations at other institutions, patients who were treated before imaging, and patients with extraaxial or entirely cystic tumors were excluded.
We retrospectively screened 35 consecutive patients to identify those who met these criteria. Twelve patients (six boys, five girls, one infant) were included in the subsequent analysis. The diagnoses were histologically confirmed in all patients. The number of each tumor type was small; therefore, we subdivided each tumor into three basic categories: low-grade (World Health Organization, grade I-II) gliomas (n = 4), nonembryonal high-grade (World Health Organization, grade III-IV) tumors (n = 4), and embryonal tumors (n = 4) (Table 1).
Age | Tumor Histology | Location | Classification | ADC Ratio | ADC Value | Cellularity | Average Nuclear Diameter (μm) | Calculated Total Nuclear Area (μm2) |
---|---|---|---|---|---|---|---|---|
8 years | Pilocytic astrocytoma | Cerebellum | Low-grade glioma | 2.50 | 1.540 | 34 | 7 | 44,506 |
9 years | Anaplastic ependymoma | Frontal lobe | High-grade other | 1.57 | 1.158 | 60 | 15 | 636,428 |
10 years | Pilocytic astrocytoma | Cerebellum | Low-grade glioma | 1.75 | 1.132 | 32 | 8 | 51,492 |
10 years | Glioblastoma multiforme | Thalamus | High-grade other | 1.45 | 1.128 | 51 | 12 | 294,284 |
7 years | Pilocytic astrocytoma | Temporal lobe | Low-grade glioma | 1.90 | 1.405 | 24 | 7 | |
2 years | Atypical teratoidrhabdoid | Frontal lobe | Embryonal | 0.70 | 0.598 | 116 | 10 | 1,057,257 |
3 months | High-grade gliomaa | Suprasellar | High-grade other | 1.59 | 1.303 | 90 | 8 | 407,314 |
16 years | Pilocytic astrocytoma | Cerebellum | Low-grade glioma | 2.54 | 1.203 | 50 | 10 | 196,428 |
13 years | Medulloblastoma | Cerebellum | Embryonal | 0.92 | 0.538 | 67 | 14 | 691,306 |
17 years | Germinoma | Suprasellar | High-grade other | 1.80 | 1.301 | 50 | 15 | 441,964 |
10 years | PNET | Parietal lobe | Embryonal | 1.26 | 0.974 | 68 | 7.5 | 204,364 |
14 years | Medulloblastoma | Cerebellum | Embryonal | 1.22 | 0.776 | 107 | 8 | 575,721 |
Note.—ADC = apparent diffusion coefficient, PNET = primitive neuroectodermal tumor. |
a
This neoplasm was indeterminate for glial cell type; differential diagnosis was anaplastic astrocytoma versus anaplastic ependymoma.
All examinations were performed on a 1.5-T system (AnalyzeAVW software; Mayo Clinic Foundation, Rochester, MN) with circularly polarized head coils. The diffusion-tensor—imaging protocol (TR/TE, 3000.0/97.4) consisted of a single-shot multisection spin-echo echoplanar pulse sequence with a 24 × 24 cm field of view, 5-mm section thickness, and a 1-mm gap between sections [3, 4]. Four tetrahedrally oriented diffusion-weighted images (diffusion sensitivity, b = 1012.4 sec/mm2), three orthogonally oriented diffusion-weighted images (b = 337.5 sec/mm2), and a reference T2-weighted intensity image (b = 0.0 sec/mm2) were obtained at each transverse section. Fourteen transverse sections were acquired in 35 sec with a 96 × 128 voxel matrix (2.50 × 1.88 × 5.00 mm voxels), interpolated to a 192 × 256 pixel matrix. All images were realigned in two dimensions, with a combination of intra- and cross-modality affine realignment procedures, to correct for image displacements and linear stretch and shear due to eddy currents [4]. For technical reasons, the diffusion-tensor images must be oriented in the transverse plane relative to the magnet bore. The diffusion-tensor images were, therefore, not necessarily acquired in register with those of the conventional MR sequences in the clinical neuroimaging protocol, which were oriented along the plane parallel to the anterior commissure and posterior commissure.
For each pixel, the elements of the diffusion tensor were derived from this combination of tetrahedral and perpendicular diffusion measurements [3, 4]. The reference T2-weighted intensity image was not included in the diffusion-tensor calculations because of the presence of artifacts in some studies that arose from spurious free induction decay signal. The apparent diffusion coefficient, a measure of the directionally averaged magnitude of diffusion with units of mm2/sec, was computed as one third of the trace of the diffusion tensor.
The MR images were analyzed as follows: we used T2-weighted images and enhanced T1-weighted images to define the slice to be analyzed, to exclude regions of hemorrhage, and to distinguish cystic from enhancing solid portions of the mass (Figs. 1A,1B,1C,2A,2B,2C,3A,3B,3C). We drew ROIs on the apparent diffusion coefficient and anisotropy maps, using AnalyzeAVW software (Mayo Clinic Foundation, Rochester, MN) in areas of solid tumor (Figs. 1A,1B,1C,2A,2B,2C,3A,3B,3C). As many as three ROIs were drawn on one or two transverse slices. Areas of edema, cystic change, and central necrosis were excluded from the analysis.









ROIs were obtained from contralateral homologous brain regions to serve as control regions. The apparent diffusion coefficient and anisotropy values were computed from the mean of the pixel values obtained for the ROIs placed on these images. When more than one region was drawn, the mean apparent diffusion coefficient and anisotropy values were calculated from all regions and averaged by the number of pixels. The ratio of the apparent diffusion coefficient to contralateral control brain region was then computed (apparent diffusion coefficient ratio) (Table 1). This computation was performed because apparent diffusion coefficient values of normal brain vary with age [5], and the contrast achieved between tumor and surrounding tissue depends on this ratio. One author drew all ROIs, and proper placement was confirmed by a consensus of a board-certified radiologist and an attending neuroradiologist with a certificate of added qualification in neuroradiology.
A board-certified attending neuropathologist performed the histopathologic analysis. Tumor cellularity was defined as the number of cells across a single representative high-power field (Fig. 4A,4B,4C,4D). We measured average cellular and nuclear diameters for each tumor, using an ocular micrometer. Total cellular area was computed from the measured cell area of each tumor cell type multiplied by the number of cells per high-power field. Total nuclear area was computed from the measured nuclear area of each tumor cell type multiplied by the number of cells per high-power field. The neuropathologist was blinded to the MR imaging findings and the apparent diffusion coefficient values for each tumor. We made statistical comparisons of apparent diffusion coefficient ratio to normal brain with tumor type, using one-way analysis of variance. Statistical comparisons of the absolute apparent diffusion coefficient values with tumor cellularity and with total nuclear area were made with linear regression analysis and Pearson's correlation coefficient. Absolute apparent diffusion coefficient values were used for this comparison to evaluate the possibility that apparent diffusion coefficient values change with these parameters. Ratios to normal brain tissue were not used in this case because it is unlikely that apparent diffusion coefficient values change with the age of the subject (i.e., with the apparent diffusion coefficient values of normal brain).




Results
The apparent diffusion coefficient values for low-grade gliomas are 1.33 ± 0.21 × 10-3 mm2/ sec (range, 1.132-1.60 × 10-3 mm2/sec) with a ratio to normal brain of 2.10 ± 0.35 (range, 1.75-2.54) (Fig. 5). The nonembryonal high-grade tumor apparent diffusion coefficients are 1.22 ± 0.09 × 10-3 mm2/sec (range, 1.128-1.303 × 10-3 mm2/sec) and ratio to normal brain, 1.6 ± 0.15 (range, 1.45-1.8) (Fig. 5). The group of embryonal tumors (primitive neuroectodermal tumor, medulloblastoma, malignant teratoid-rhabdoid tumor) have apparent diffusion coefficient values of 0.72 ± 0.20 × 10-3 mm2/sec (range, 0.538-0.974 × 10-3 mm2/sec) and a ratio to normal brain of 1.03 ± 0.26 (range, 0.7-1.26) (Fig. 5). There is a clear distinction between the low-grade gliomas and the embryonal tumors, with the high-grade nonembryonal tumor group (glioma, germ cell, and ependymoma) having some overlap (p = 0.001). Apparent diffusion coefficient values are inversely correlated with tumor cellularity (Fig. 6); the apparent diffusion of water is slower in tumors with higher cellularity. This relationship is statistically significant (p < 0.05). In addition, apparent diffusion coefficient is inversely correlated with the estimated total nuclear area (p < 0.01) (Fig. 7). The anisotropy values were decreased to near zero in all tumors and, thus, were not analyzed further. Total cellular area was computed and compared with the apparent diffusion coefficient values, but this correlation did not reach statistical significance in this study.



Discussion
Diffusion imaging has been widely used in the evaluation of acute cerebral ischemia [6, 7]. The mechanism of the decrease in apparent diffusion coefficient observed with recent stroke is controversial [8, 9], but imaging findings are quite reliable. Diffusion imaging is also useful in distinguishing tumor from abscess [10,11,12,13]. These studies have shown that the apparent diffusion coefficient values of abscess fluid are markedly lower than those of cystic or necrotic portions of tumor. This difference is thought to be due to the impeded water mobility of purulent fluid related to its high cellularity and viscosity [13]. Diffusion imaging has also been shown to distinguish epidermoid from arachnoid cysts [14, 15]. Arachnoid cysts are characterized by free diffusion similar to that of cerebrospinal fluid. Epidermoids, conversely, show slower diffusion, presumably as a result of the more complex nature at the microscopic level, which impedes water diffusion. These observations prompted this investigation of the relationship of the apparent diffusion coefficient of water to the complexity of the tissue sample at the microscopic level.
In our study, the apparent diffusion coefficient correlated significantly with both the tumor cellularity and the calculated total nuclear area. The apparent diffusion coefficient of the 13-year-old girl with medulloblastoma was seen as an outlier in the tumor cellularity graph. This feature can likely be explained by the large cell pathology of this medulloblastoma. This patient had a high calculated total nuclear area but had lower cellularity (Fig. 4D). This tumor displayed restricted diffusion presumably because of the densely packed nature of the tumor cells and the high nuclear-to-cytoplasm ratio.
To our knowledge, published data characterizing diffusion in pediatric brain tumors are limited. Several studies have shown that different components of tumor, surrounding edema, and normal surrounding white matter have different apparent diffusion coefficient values [16,17,18,19,20,21]. Cystic and necrotic portions of tumor have higher apparent diffusion coefficient values, reflecting the increased water movement characterizing these components. Tien et al. [17] used diffusion-weighted imaging along a single axis to show that enhancing components of highgrade gliomas have a lower apparent diffusion coefficient than nonenhancing tumor and peritumoral edema. Brunberg et al. [19] found a distinction between apparent diffusion coefficient and diffusion anisotropy from normal white matter and solid enhancing tumor, cystic and necrotic areas, and regions of edema. They did not find a difference in apparent diffusion coefficient values between various glioma subtypes. Sugahara et al. [2] showed a trend similar to our study by comparing high- and low-grade gliomas to cellularity defined as the total area of nuclei of tumor cells divided by the area of histologic section. They concluded that apparent diffusion coefficient correlated with tumor cellularity.
A previous case report postulated that the densely cellular nature and high nuclear-to-cytoplasmic ratio of medulloblastoma restricts extracellular diffusion of water [22]. However, no quantitative data were presented to support this hypothesis. Our data support this hypothesis by showing that these tumors have apparent diffusion coefficient values of less than 1.0 × 10-3 mm2/sec, consistent with reduced diffusion.
Chenevert et al. [23] used a rat—brain tumor model to show that diffusion measurements are sensitive to chemotherapy induced changes. Apparent diffusion coefficient values were obtained pretreatment, after treatment, and during tumor regrowth. Results indicated that the apparent diffusion coefficient values increase with therapy and return to the pretreatment levels during tumor regrowth. The increase in apparent diffusion coefficient during therapy suggests an increase in interstitial volume and cellular permeability from tumor-cell necrosis. Further investigation along these lines could lead to new ways of evaluating tumor response to therapy.
Our study is limited by the small number of patients and by the heterogeneous sample of tumors. ROIs analyzed on MR images may not precisely correspond to the biopsy specimens used in the pathologic evaluations. Finally, our values for the total nuclear area are calculated estimates and are not actual measurements. Despite these limitations, we believe that the observations are potentially clinically significant. All studies were performed preoperatively, and thus, there were no influences of surgery or prior therapy. Further studies in this pediatric population are needed to determine the role of this modality in clinical decision-making.
In conclusion, we found that apparent diffusion coefficients of pediatric brain tumors are significantly correlated with tumor cellularity and total nuclear area. Quantitative diffusion imaging may enhance the diagnostic process and may be predictive of patient clinical outcome before tissue sampling of pediatric CNS malignancies.
Acknowledgments
We thank Thomas K. Pilgram for expert advice on the statistical analyses used in this investigation.
Footnotes
Supported by the American Academy of Pediatrics Resident Research Grant.
Address correspondence to K. M. Gauvain.
References
1.
Becker L. Pathology of pediatric brain tumors. Neuroimaging Clin N Am 1999; 9:671-690
2.
Sugahara T, Korogi Y, Kochi M, et al. Usefulness of diffusion-weighted MRI with echo-planar technique in the evaluation of cellularity in gliomas. J Magn Reson Imag 1999; 9:53-60
3.
Mukherjee P, Bahn MM, McKinstry RC, et al. Differences between gray matter and white matter water diffusion in stroke: diffusion-tensor MR imaging in 12 patients. Radiology 2000; 215:211-220
4.
Shimony JS, McKinstry RC, Akbudak E, et al. Quantitative diffusion-tensor anisotropy brain MR imaging: normative human data and anatomic analysis. Radiology 1999; 212:770-784
5.
Mukherjee P, Miller JH, Shimony JS, et al. Normal brain maturation during childhood: developmental trends characterized with diffusion tensor MR imaging. Radiology 2001 (in press)
6.
Gray L, MacFall J. Overview of diffusion imaging. Magn Reson Imaging Clin N Am 1998; 6:125-138
7.
Rowley HA, Grant PE, Roberts TPL. Diffusion MR imaging. Neuroimaging Clin N Am 1999; 9:343-361
8.
Beauchamp NJ, Ulug AM, Passe TJ, van Zijl PC. MR diffusion imaging in stroke: review and controversies. RadioGraphics 1998; 18:1269-1283
9.
Duong TQ, Ackerman JJ, Ying HS, Neil JJ. Evaluation of extra- and intracellular apparent diffusion in normal and globally ischemic at brain via 19f NMR. Magn Reson Med 1998; 40:1-13
10.
Kim YJ, Chang KH, Song IC, et al. Brain abscess and necrotic or cystic brain tumor: discrimination with signal intensity on diffusion-weighted MR imaging. AJR 1998; 171:1487-1490
11.
Ebisu T, Tanaka C, Umeda M, et al. Discrimination of brain abscess from necrotic or cystic tumors by diffusion-weighted echo planar imaging. Magn Reson Imaging 1996; 14:1113-1116
12.
Noguchi K, Watanabe N, Nagayoshi T, et al. Role of diffusion-weighted echo-planar MRI in distinguishing between brain abscess and tumor: a preliminary report. Neuroradiology 1999; 41:171-174
13.
Desprechins B, Stadnik T, Koerts G, Shabana W, Breucq C, Osteaux M. Use of diffusion-weighted MR imaging in differential diagnosis between intracerebral necrotic tumors and cerebral abscesses. AJNR 1999; 20:1252-1257
14.
Laing AD, Mitchell PJ, Wallace D. Diffusion-weighted magnetic resonance imaging of intracranial epidermoid tumours. Australas Radiol 1999; 43:16-19
15.
Tsuruda JS, Chew WM, Moseley ME, Norma D. Diffusion-weighted MR imaging of the brain: value of differentiating between extraaxial cysts and epidermoid tumors. AJR 1990; 155:1059-1065
16.
Krabbe K, Gideon P, Wagn P, Jansen U, Thomsen C, Madsen F. MR diffusion imaging of human intracranial tumors. Neuroradiology 1997; 39:483-489
17.
Tien RD, Felsberg GJ, Friedman H, Brown M, MacFall J. MR imaging of high-grade cerebral gliomas: value of diffusion-weighted echoplanar pulse sequences. AJR 1994; 162:671-677
18.
Eis M, Els T, Hoehn-Berlage M. Quantitative diffusion MR imaging of cerebral tumor and edema. Acta Neurochir (Wien) 1994; 60:344-346
19.
Brunberg JA, Chenevert TL, McKeever PE, et al. In vivo MR determination of water diffusion coefficients and diffusion anisotropy: correlation with structural alteration in gliomas of the cerebral hemispheres. AJNR 1995; 16:361-371
20.
Le Bihan D, Douek P, Argyropoulou M, Turner R, Patronas N, Fulham M. Diffusion and perfusion magnetic resonance imaging in brain tumors. Top Magn Reson Imaging 1993; 5:25-31
21.
Le Bihan D, Turner R, Douek P, Patronas N. Diffusion MR imaging: clinical applications. AJR 1992; 159:591-599
22.
Kotsenas AL, Roth TC, Manness WK, Faeber EN. Abnormal diffusion-weighted MRI in medulloblastoma: Does it reflect small cell histology? Pediatr Radiol 1999; 29:524-526
23.
Chenevert TL, McKeever PE, Ross BD. Monitoring early response of experimental brain tumors to therapy using diffusion magnetic resonance imaging. Clin Cancer Res 1997; 3:1457-1466
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: December 1, 2000
Accepted: February 16, 2001
First published: November 23, 2012
Authors
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.