Differentiation Between Classic and Atypical Meningiomas with Use of Diffusion Tensor Imaging

BACKGROUND AND PURPOSE:The differentiation between classic and atypical meningiomas may have implications in preoperative planning but may not be possible on the basis of conventional MR imaging. Our hypothesis was that classic and atypical meningiomas have different patterns of intratumoral water diffusion that will allow for differentiation between them. MATERIALS AND METHODS:Preoperative diffusion tensor imaging (DTI) was performed in 12 classic and 12 atypical meningiomas. Signal intensity of solid-enhancing tumor regions on diffusion-weighted trace images and apparent diffusion coefficient (ADC) and fractional anisotropy (FA) maps was assessed. Regions of interest (ROIs) were placed in solid-enhancing regions, peritumoral edema, and contralateral normal-appearing white matter (NAWM) to measure tensor metrics including major (λ1), intermediate (λ2) and minor eigenvalues (λ3) and FA and ADC values. Distribution of tensor shapes within enhancing tumors was calculated for all tumors. Differences between classic and atypical meningiomas in tumor signal intensity, intratumoral and peritumoral tensor metrics, as well as tensor shapes distribution were statistically analyzed. RESULTS:A significantly greater proportion of atypical meningiomas were isointense and hypointense on ADC maps (P = .007). Classic meningiomas had significantly lower FA (P = .012), higher ADC (P = .011), greater λ2 (P = .020) and λ3 (P = .003). There was significantly more spherical diffusion in classic than in atypical meningiomas (P = .020). All diffusion tensor metrics for peritumoral edema of the 2 tumor groups did not differ. CONCLUSION:DTI showed that intratumoral microscopic water motion is less organized in classic than in atypical meningiomas. This feature may allow for noninvasive differentiation between classic and atypical meningiomas.

M eningiomas account for between 16% and 20% of primary intracranial tumors. 1 According to the World Health Organization (WHO) classification system, 78% of meningiomas are grade I, 20.4% are grade II, and 1.6% are grade III. 2 Grades II and III meningiomas are more aggressive than grade I meningiomas. Five-year recurrence rates are 12% for benign meningiomas and 41% for atypical meningiomas. 2 Initial extent of tumor resection and histologic grade are key determinants for recurrence. 3 Therefore, prospectively identifying their histologic grades can be clinically beneficial in treatment planning. Although conventional MR imaging can provide detailed morphologic information of meningiomas, its value in the prediction of WHO grades is limited. 4 According to the WHO classification, 5 classic meningiomas differ from atypical ones in their number of mitoses, cellularity, and nucleus-to-cytoplasm ratio (N/C ratio) as well as their histologic patterns. Complex microstructural barriers in brain tissue, such as white matter tracts, cell membranes, and capillary vessels result in a tendency for water molecules to diffuse with direction (anisotropic diffusion) rather than equally in all directions (isotropic diffusion). Isotropic diffu-sion-weighted imaging (DWI), which measures average magnitude of water motion in apparent diffusion coefficient (ADC), has shown controversial results for differentiating classic from atypical meningiomas. [6][7][8] In contrast to isotropic DWI, diffusion tensor imaging (DTI) provides information about magnitude and directionality of water diffusion 9 and thus may be able to measure the differences in intratumoral diffusion anisotropy as a result of histologic differences between classic and atypical meningiomas. On the other hand, peritumoral edema associated with meningiomas, regardless of classic or atypical subtypes, has always been considered a purely vasogenic edema (ie, absence of tumor cell infiltration). 10,11 Our first hypothesis was that intratumoral diffusion anisotropy is different between these 2 tumor types and that differences in diffusion anisotropy as detected by DTI allow differentiation between them. Our second hypothesis was that anisotropic diffusion measured in peritumoral edema with DTI will not be different between classic and atypical meningiomas.

Patients
We obtained approval for this study from the Institutional Board of Research Associates and obtained signed informed consent from all patients. Patient identifiers were removed from image data before analysis. No patients had begun corticosteroid treatment or radiation therapy, and none had previous brain biopsy at the time of MR imaging. MR examinations were performed in 12 patients (7 men, 5 women; mean age, 55.2 years; age range, 38 -71 years) with classic meningiomas (WHO grade I) and 12 patients (5 men, 7 women; mean age, 56.3 years; age range, 24 -80 years) with atypical meningiomas (WHO grade II). Histologic diagnosis was obtained in all patients. Subtypes of meningiomas were classified according to the WHO (2007) classification. 5 An atypical meningioma has increased mitotic activity, which is defined as 4 or more mitoses per 10 highpower fields or with 3 or more of the following features: 1) increased cellularity, 2) small cells with a high nucleus-to-cytoplasm ratio, 3) prominent nucleoli, 4) uninterrupted patternless or sheetlike growth, and 5) foci of "spontaneous" or "geographic" necrosis. Subtypes of classic meningiomas included 7 meningothelial, 2 fibroblastic, 2 transitional, and 1 psammomatous meningioma. Peritumoral edema was present in 9 (75%) of 12 patients with classic meningiomas and 8 (66.6%) of 12 patients with atypical meningiomas. Tumors with large calcifications, hemorrhages, or both were excluded.

MR Imaging
All patients underwent MR imaging including axial T2-weighted, DWI, DTI, and axial spin-echo T1-weighted imaging (T1WI) performed before and after intravenous administration of 0.1 mmol/kg body weight of gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany). All MR imaging studies were performed on a single occasion with a 3T unit (Magnetom Tim Trio; Siemens, Erlangen, Germany). DTI was performed in the axial plane with single-shot echo-planar imaging with the following parameters: TR, 8600 ms; TE, 91 ms; diffusion gradient encoding in 12 directions; b ϭ 0; 1000 seconds/mm 2 ; FOV, 192 ϫ 192 mm; matrix size, 128 ϫ 128; section thickness, 3 mm; and number of signal intensity acquired, 5. A total of 35 to 40 sections were used to cover the cerebral hemispheres, upper brain stem, and cerebellum without gaps. To minimize artifacts such as signal intensity drop-out and gross geometric distortions associated with the echo-planar imaging, we used the parallel imaging technique (generalized autocalibrating partially parallel acquisitions 12 with a reduction factor of 2) during DTI acquisitions.

Image Postprocessing
We transferred diffusion tensor data to an independent workstation and processed using the software nordicICE (nordic Image Control and Evaluation Version 2.2; Nordic Imaging Lab, Bergen, Norway). The diffusion tensor was diagonalized to yield the major ( 1 ), intermediate ( 2 ), and minor ( 3 ) eigenvalues. Voxel by voxel, the fractional anisotropy (FA), directionally averaged ADC (also known as mean diffusivity), and distribution of linear (C l ), planar (C p ), or spherical (C S ) tensor shapes 13 were calculated with the following respective standard algorithms:

Imaging Analysis
Two neuroradiologists performed qualitative visual inspections of DW trace images and ADC and FA maps with consensus reading. On DW trace images and ADC maps, each lesion was categorized as being predominantly hyperintense, isointense, or hypointense relative to the cortex. On FA maps, signal intensity was judged as predominantly hypointense if similar to that of the cortex, hyperintense if similar to white matter, and otherwise as mixed signal intensity. With use of the coregistration module integrated of nordicICE, the 1 , 2 , 3 , and FA and ADC maps were automatically co-registered to postcontrast T1WI on the basis of DICOM geometry parameters. The adequacy of registration was visually assessed by the 2 observers, and manual adjustments were performed if necessary. Circular regions of interest (ROIs) with diameters ranging from 1 to 1.5 cm were placed centrally within the largest solid-enhancing area of all meningiomas and peritumoral area if edema was present (Figs 1 and 2). ROIs were then automatically transferred to the co-registered 1 , 2 , 3 , and FA and ADC maps. ROIs were copied onto the corresponding contralateral normal-appearing white matter (NAWM) in each patient to obtain the 1 , 2 , 3 , and FA and ADC values for the purpose of normalization. The neuroradiologist placing the ROIs was blinded to the individual diagnoses. For each ROI within the enhancing tumor, the distribution of tensor shapes-categorized as linear, planar, and spherical-was calculated.

Statistical Analysis
Differences in signal intensities on DW trace images and ADC and FA maps between classic and atypical meningiomas were compared with a 2 test. Mean 1 , 2 , 3 , and FA and ADC values were evaluated for each ROI. The ratios of 1 , 2 , 3 , and FA and ADC were calculated by dividing 1 , 2 , 3 , and FA and ADC values in the solid-enhancing tumor areas or peritumoral edema, if present, by 1 , 2 , 3 , and FA and ADC values of the contralateral NAWM in the same patient. Comparisons between solid-enhancing areas of classic and atypical meningiomas and corresponding contralateral NAWM were performed with use of paired t tests. Mean absolute values and ratios of 1 , 2 , 3 , and FA and ADC of solid-enhancing areas and peritumoral edema, as well as the distribution of tensor shapes of the solid-enhancing areas for the 2 tumor types were compared with a 2-sample t test. A commercially available statistical software package (SPSS 15; SPSS, Chicago, Ill) was used for analysis, and P values less than .05 were considered to indicate statistically significant differences.

Qualitative Visual Inspections of DW Trace Images and ADC and FA Maps
Results of qualitative comparison of DW trace images and ADC and FA maps between classic and atypical meningiomas are summarized in Table 1. There were 9 (75%) of 12 classic meningiomas that were hyperintense and 3 (25%) of 12 that were isointense or hypointense on DW trace images. All 12 atypical meningiomas were hyperintense on DW trace images. Differences in signal intensity on DW trace images between classic and atypical meningiomas were not significant (P ϭ .064). On ADC maps, 9 of 12 classic meningiomas were hyperintense, 3 (25%) of 12 were isointense; none was hypointense. On ADC maps, 2 (16.6%) of 12 atypical meningiomas were hyperintense, 5 (41.7%) of 12 were isointense, and 5 (41.7%) of 12 were hypointense. Differences in signal intensities on ADC maps between classic and atypical meningiomas were significant (P ϭ .007). One (8.3%) classic meningioma was hyperintense, 8 (66.7%) of 12 had mixed signal intensity, and 3 (25%) of 12 were hypointense on FA maps. There were 3 (25%) of 12 atypical meningiomas that were hyperintense, 8 (66.7%) of 12 that had mixed signal intensities, and 1 (8.3%) of 12 that was hypointense on FA maps. Differences in signal intensities on FA maps between classic and atypical meningiomas were not significant (P ϭ .368).

Mean FA Values and FA Ratios
Results of quantitative comparison intratumoral and peritumoral FA between classic and atypical meningiomas are summarized in Tables 2 and 3

Mean ADC Values and ADC Ratios
Results of quantitative comparison of intratumoral and peritumoral ADC between classic and atypical meningiomas are summarized in Tables 2 and 3

Mean Values and Ratios of 1 , 2 , 3
Results of quantitative comparison of intratumoral and peritumoral eigenvalues between classic and atypical meningiomas are summarized in Tables 2 and 3 Mean values and ratios of 2 and 3 of classic meningiomas were significantly greater than those of atypical meningiomas. Mean 1 values and ratios, on the other hand, were not significantly different between the 2 types of tumors. The peritumoral edema of classic and atypical meningiomas was not significantly different in mean 1 , 2 , and 3 values and ratios.

Tensor Shapes Assessment
Results of intratumoral tensor shape distributions in classic and atypical meningiomas are summarized in Table 4. Mean percentages of linear, planar, and spherical diffusion were 8.1% Ϯ 3%, 14.7% Ϯ 8%, and 77.2% Ϯ 10%, respectively, for classic meningiomas and 13% Ϯ 8%, 19.4% Ϯ 8%, and 67.6% Ϯ 9%, respectively, for atypical meningiomas. There was significantly more spherical diffusion in classic meningiomas than in atypical meningiomas (P ϭ .020). The proportions of linear and planar diffusion were higher in atypical meningiomas compared with classic meningiomas, but these differences were not statistically significant.

Discussion
Our study shows that classic and atypical meningiomas cannot be differentiated on the basis of their signal intensity on DW trace images and FA maps. However, they could be differentiated with ADC maps because atypical meningiomas were generally isointense or hypointense, whereas classic meningiomas were generally hyperintense on ADC maps. On quantitative analysis, solid-enhancing areas of classic meningiomas had  significantly lower FA, higher ADC, greater 2 and 3 values, and a greater proportion of spherical tensors compared with atypical meningiomas. Greater diffusivity along transverse axes ( 2 and 3 ) of the diffusion tensor as well as a greater proportion of spherical tensors indicate that microscopic water motion is more disorganized in classic than in atypical meningiomas. On the other hand, quantitative DTI analysis of peritumoral edema of the 2 tumor groups did not differ. On microscopic examination, classic meningiomas consist of oval or spindle-shaped neoplastic cells that form whorls, fascicles, cords, or nodules. Histologic subtypes of classic meningiomas are based on tumoral architecture and cytologic features. Among these, the meningothelial, fibrous, and transitional variants are considered the prototypical histologic patterns of classic meningiomas. 14 A single meningioma may show a combination of meningothelial, transitional, and fibrous histologic patterns. Other variants of classic meningiomas, despite having other architectural and cytologic features, are usually combined with the 3 prototypical histologic patterns. 14 We speculate that the whorls, fascicles, cords, or nodules found in classic meningiomas may serve as microscopic physical barriers and prevent water molecules from moving in a linear fashion. Consequently, microscopic water movement in classic meningiomas lacks a coherent organization throughout the tumor that results in increased 2 and 3 as well as a greater proportion of spherical tensors. In contrast to classic meningiomas, histologic features of atypical meningiomas are often described as uninterrupted patternless or sheetlike growths. Sheetlike growth pattern refers to a lack of any specific architectural pattern within the tumor. Neoplastic cells in atypical meningiomas do not form widespread whorls, fascicles, cords, or nodules as they do in classic meningiomas. Thus, we speculate that water molecules tend to move with more directionality in atypical meningiomas because of the absence of these microscopic physical barriers. This pattern is reflected as higher diffusion anisotropy and less spherical diffusion tensors in atypical meningiomas compared with classic ones.
Besides uninterrupted patternless or sheetlike growth pattern, other histologic features of atypical meningiomas on the basis of WHO criteria include 1) 4 or more mitoses per 10 high-power fields, 2) increased cellularity, 3) small cells with a high N/C ratio, 4) prominent nucleoli, and 5) foci of spontaneous necrosis. 5 Among these, cellularity and N/C ratio of tumors correlate inversely with water diffusivity. 15 In our study, the number of tumors that were isointense or hypointense was significantly greater in atypical than in classic meningiomas (83.4% vs 25%) on the basis of qualitative analysis of ADC maps. On quantitative analysis, mean ADC values and ratios were significantly lower in atypical than in classic meningiomas. These findings reflect the greater cellularity and higher N/C ratio found in atypical meningiomas. Although   results from earlier studies on whether ADC can differentiate classic from atypical meningiomas are controversial, 6-8 the results from our study show that the ADC of these 2 tumors are different. On the other hand, although all atypical meningiomas were hyperintense on DW trace images, 75% of classic meningiomas also were hyperintense. Similar to a previous study, 8 our study shows that DW trace images cannot be used to differentiate atypical meningiomas from classic meningiomas.
A previous study 8 has found no difference in ADC between peritumoral edema of classic and atypical meningiomas. In our study, quantitative measurements of both isotropic and anisotropic water diffusion demonstrated no differences between peritumoral edema of both tumor groups. Brain invasion occurs in both classic (43%) and atypical (57%) meningiomas. 16 Although peritumoral edema of meningiomas has been proposed to be related to brain invasion, 17 it can occur without brain invasion and its pathogenesis is also related to the size of the tumor, histologic subtype, vascularity, secretory activity, tumor-related venous obstruction, and expression of sex hormones and receptors. 18 Furthermore, in meningiomas with brain invasion, the invasive neoplastic cells infiltrate only the cerebral cortex and generally do not extend into white matter. 17 Histologic analysis of peritumoral brain tissue of meningiomas with light and electron microscopy has shown only extracellular fluid accumulation. The neurons and fibers were relatively well preserved. 11 Vasogenic edema found in both tumor groups likely accounted for the absence of differences on DTI. Therefore, analysis of the diffusion anisotropy of peritumoral edema for differentiation between classic and atypical meningiomas with DTI is not useful.
Limitations of our study included, first, that the effect of microscopic tumor growth patterns on water diffusion was not quantified. Therefore, we could not definitively state that differences in diffusion anisotropy between classic and atypical meningiomas were only because of differences in tumor architecture. Second, we did not compare the cellularity in classic and atypical meningiomas, and thus we could not definitively state that the ADC differences between these 2 tumors were because of differences in cellularity. Third, we did not study other less common subtypes of WHO grade I meningiomas such as angiomatous, microcystic, secretory, and metaplastic meningiomas. From a histologic standpoint, these less common meningiomas do form specific architectural patterns. Therefore, we speculate that their diffusion anisotropy would be different from atypical meningiomas that are "featureless" on microscopic examination. In a similar fashion, other WHO grade II meningiomas (ie, chordoid and clear cell meningiomas) were not encountered during our study period.

Conclusion
In our preliminary study, classic meningiomas had significantly lower intratumoral FA, higher ADC, and greater 2 and 3 as well as a greater proportion of spherical tensors compared with atypical meningiomas. These DTI findings indicate that intratumoral microscopic water motion is more disorganized in classic than in atypical meningiomas. DTI may be helpful in the differentiation between classic meningiomas and atypical meningiomas.