Distinguishing between Germinomas and Pineal Cell Tumors on MR Imaging

Can one distinguish germinomas from pineal cell tumors or do they look the same on MRI? ADC values, T1-, and T2-weighted images from 20 patients with preoperative pineal tumors were assessed in an attempt to differentiate between these 2 tumors. The only significant difference was higher ADC values for germinomas. T1 and T2 images were not capable of differentiating between these tumors. Patients with germinomas were younger than those with pineal cell tumors (pineoblastomas, papillary tumor, and intermediate tumor). BACKGROUND AND PURPOSE: Tumors of pineal cell origin have different prognosis and treatment than those of germ cell origin. The recent literature suggests that these tumors often look alike. Our study aimed to differentiate between pineal cell tumor and germinoma based on ADC values, the homogeneity of the mass, and MR imaging characteristics. MATERIALS AND METHODS: We enrolled 20 patients who had pretreatment MR imaging scans with histologic verification of tumors of pineal cell origin and germinomas. The tumors were measured for the ADC values and for homogeneity by the coefficient of variation of ADC values, and T1WI and T2WI signal intensity values. RESULTS: The 20 subjects (8 females and 12 males) ranged in age from 1.5–64.9 years, with a mean age of 23.9 years (SD 17.7 years). The mean age of those with germinomas was 13.7 years (SD 3.8 years), less than the mean of 29.4 years for those with pineal cell tumors (SD 19.9 years; P = .016). These 2 groups showed no significant difference in coefficients of variation on T1WI, T2WI, and ADC images. However, germinomas showed statistically significant higher ADC values (mean 1590.69 ± 532.96 × 10−6 mm2/s) than pineal cell tumors (mean 883.58 ± 317.48 × 10−6 mm2/s; P = .02). An accuracy of 89.5%, sensitivity of 83.3%, specificity of 92.3%, PPV of 83.3%, and NPV of 92.3% were yielded for an ADC threshold of 1250.00 × 10−6 mm2/s. CONCLUSIONS: Germinomas showed higher ADC values than the pineal cell tumors (P = .02), and the patients were younger. Otherwise, there were no definitive imaging characteristics that distinguished pineal cell tumors from germinomas.

P ineal region tumors constitute 1% of all intracranial tumors 1 and account for 3%-8% of intracranial tumors in children. 2 Among the various histologic types of pineal tumors, germinomas are the most frequent, followed by teratomas. The germ cell tumors constitute approximately 35% 3 of all pineal region tumors. The pineal cell origin tumors, that is, pineoblastoma and pineocytoma, are less frequently seen (approximately 28% of the total) 3 and are believed to be differentiated from other cell types by their characteristic dispersion rather than engulfment of the pineal calcifications, as seen in germinomas. The imaging literature suggests that the common pineal region tumors have no pathognomonic imaging patterns. 4 In addition, approximately 11% of biopsies are either nondiagnostic or misdiagnosed 5 due to the complexity of the masses and their high vascularity, leading to insufficient tissue.
Because of the overlapping imaging findings on conventional techniques, scanning may be less useful than serum markers for differential diagnosis specificity. DWI is an imaging technique that has been applied for evaluation of cerebral infarction, 6,7 glioma and lymphoma, [8][9][10] and brain abscesses 11,12 because it provides image contrast based on molecular (Brownian) motion of water. 6,13 ADC values generated from DWI can imply the relative intracellular and extracellular volumes within the tissue being studied. [8][9][10] However, information on the diffusivity of pineal tumors is limited. With the wide clinical availability of DWI, measurements of ADC values are more frequently integrated into conventional imaging analysis and may help differentiate between germinomas and pineal cell tumors. In addition, we have anecdotally observed that germinomas tend to be more homogeneous in MR signal intensity than other tumors.
Because there is a difference between the prognosis and treatment of tumors of pineal cell origin and those of germ cell origin, our study aimed to differentiate between pineal cell tumors and germinomas based on ADC values, the homogeneity of the mass, and other MR imaging characteristics.

Materials and Methods
We enrolled 20 patients with histologic verification of a pineal cell origin tumor and germinoma from our institute by using query keywords of the Radiology Information System and pathology data base for "pineoblastoma," "pineocytoma," and "germinoma" during a query period from 2002-2010. In this study we excluded patients with previous surgical resection or histopathologic reports outside our hospital. Demographic data were recorded from the electronic medical record. MR imaging studies were retrospectively reviewed and measured for the ADC values and for homogeneity by measuring the standard deviation divided by the means of ADC values, T1WI, and T2WI, measured by a standard ROI analysis from PACS. Specifically, we used elliptical ROIs that included the entire tumor. We then assessed the percentage of solid to cystic components of the mass by dividing the area of the cystic portion by the area of the entire mass on the axial scan that showed the greatest dimension of the mass.
The scans were also evaluated qualitatively for enhancement degree (mild [faint], moderate [intermediate], and marked [as much as expected for a meningioma] based on observation) and pattern. A diagnosis was also predicted based on overall nonquantitative imaging characteristics by the most experienced (22 years) neuroradiologist author, blinded to the age of the patient.

Imaging Techniques
The MR studies were performed with different MR scanners: 15 with a 1.5T (Signa HDx; GE Healthcare, Milwaukee, Wisconsin), 1 with a 3T Trio, and 4 with a 1.5T Avanto scanner (both Siemens Medical Systems, Erlangen, Germany). Nineteen patients were scanned using a 1.5T magnet, and 1 with a 3T magnet. In addition to axial DWI, conventional T1WI, T2WI, FLAIR, and postgadolinium images were performed. Three neuroradiologists retrospectively reviewed the images without information of the pathologic results.

Data Analysis
The ADC images were obtained by using postprocessing software for each vendor's MR scanner. The ADC values were measured from the RadSuite Advanced Visualization PACS (Emageon, Birmingham, Alabama). ROIs were thoroughly positioned around the border of the tumors by 2 radiologists independently in separate sessions, blinded to the final diagnosis. The regions of interest used an ellipital shape that was deformed to include the entire lesion on the axial scans to match the ADC maps, which were provided in the axial plane (Fig 1). These recorded values included the standard deviations and mean values measured on T1WI, T2WI, and ADC sequences, with heterogeneity assessed by using the coefficient of variation (SD/mean). The entire tumor-consisting of the solid, cystic, and calcified areas-was included in the ROI. The lesions were evaluated for cystic component as the portion of the tumor that did not enhance but that had fluid signal intensity on the other pulse sequences. The area of maximal cystic component on the axial scan was divided by the maximal overall area on the axial scans for the tumor to determine percentage cystic. If there was Ͼ20% difference in the measurements of ADC, T1WI, or T2WI values between 2 primary readers, a third independent radiologist remeasured the lesions, blinded to the final diagnosis, to establish the consensus value, which was averaged with its nearest neighbor to create a final mean value. Enhancement characteristics were also recorded as homogeneous or heterogeneous enhancement, and mild, moderate, and marked in degree as described.

Statistical Analysis
Statistical analyses were performed with SPSS software version 18 (SPSS, Chicago, Illinois). The study of age and coefficient of variation of the mean signal intensity values on T1WI, T2WI, and ADC between germinomas and pineal parenchymal tumors were evaluated by using the t test for independent samples, which was also applied to calculate whether these tumor groups showed significant differences according to different ADC values. A P value of less than 0.05 was considered to indicate a significant difference. Qualitative description of the enhancement degree and pattern were reported as a percentage.

Results
The 20 subjects (8 females and 12 males) ranged in age from 1.5-64.9 years, with a mean age of 23.9 years (SD 17.7 years). The mean age of patients with germinomas was 13.7 years (SD 3.8 years), less than the mean of 29.4 years for patients with pineal cell tumors (SD 19.9 years; P ϭ .016; Table 1). All pa-tients with germinoma were less than 21 years old. The histopathologic diagnoses were pineocytoma (3 cases), pineoblastoma (6 cases), germinoma (7 cases), intermediate pineal parenchymal tumors (2 cases), and papillary tumors (2 cases). The size of the germinomas ranged from maximal dimension of 0.8 to 3.8 cm, with a mean of 2.64 (SD 1.01). The pineal cell tumors ranged in maximal dimension from 1.7 to 6.1 cm, with a mean of 2.85 cm (SD 1.3). The differences in sizes were not statistically significant.
Diffuse tumoral calcification was present in only 1 case, a pineoblastoma, which showed an ADC value of 622.5, the fourth lowest in the sample. Two tumors had a punctate area of 2-mm calcification in a much larger mass. No tumors showed preoperative hemorrhage ( Table 2).
The experienced neuroradiologist gave the correct categorization of the pineal parenchymal tumors and germinomas based on conventional image features in 11/20 (55% accuracy).

Discussion
The spectrum of the pineal parenchymal tumors includes pineocytoma, PTPRs, PPTIDs, and pineoblastoma. Pineocytomas (WHO I) constitute 14%-30% of all pineal parenchymal tumors and can present in all ages, but most commonly present in adults from the third to sixth decades of age. 3 Cysts and hemorrhage may be present but necrosis is rare. 14 Gross total resection has been set as the standard treatment of this tumor, with favorable outcome. 3 CSF dissemination rarely occurs 2 and recurrences are uncommon. 3 PPTIDs (WHO II or III) constitute of 20%-62% of pineal parenchymal tumors and occur mostly in adults, with slight female preponderance. 3,14 Even though mitotic count is used to differentiate grades of PPTID, 3 definite criteria for lowgrade and high-grade PPTID have not yet been established by the WHO. 14 Management for high-grade PPTID is a combination of surgery, radiation, and chemotherapy, whereas the benefit of chemoradiation in low-grade PPTID is unclear. 3 CSF dissemination is more common in high-grade (36%) than in low-grade (7%) PPTID. 15 Pineoblastoma (WHO IV) is the most malignant type of the pineal-origin tumors; this type occurs commonly in children and constitutes 24%-50% of pineal parenchymal tumors. The histology of this tumor shows highly cellular, densely packed, patternless sheets of small cells with round to irregular nuclei and scant cytoplasm (high nuclear/cytoplasmic ratio). 14 Extensive cystic change is rare. Treatment usually includes surgery, radiation, and chemotherapy. 3 A high risk of leptomeningeal seeding was reported (45%). 16 PTPR can occur in both children and adults of a wide age range, from 5-66 years. Biologic behavior may lead to WHO classification of grade II or III, though definite criteria have yet to be determined. 3,14 Histologically, this tumor is more densely cellular, and necrotic foci are often seen. 14 No standard treatment has been established. 3 A 7% rate of CSF dissemination has been reported. 17 Germinomas account for 1%-2% of all cranial neoplasms.    Germinomas of the pineal region account for 50%-65% of intracranial germinomas, with male preponderance. Most of the patients are less than 20 years old. 2 The histology shows large undifferentiated cells that resemble primordial germinal elements, prominent round vesicular, and centrally positioned nuclei with relative abundant cytoplasm. Necrosis is uncommon. 14 Based on the abundance of cytoplasm, one might expect the ADCs of germinomas to be higher than those of the more densely cellular pineal cell tumors that have higher nuclear to cytoplasm ratios, which we found in our MR imaging.
As compared with the literature review, we found variable age distributions in both groups: 39.6 -64.9 years in pineocytomas, 45.6 -49.2 years in PPTIDs, 1.5-31.6 years in PTPRs, 3.3-45.8 years in pineoblastomas, and 9.3-20.6 years in germinomas. Our results showed a statistically significant age difference between the patients with germinomas (mean 13.7 years) versus pineal parenchymal tumors (29.4 years). Using a cutoff age of 21 years would have yielded a correct diagnosis of germinoma in all 7 patients, but would have been right in only 8 of 13 (61.5%) of pineal cell tumors. We also found male preponderance in germinomas (6/7) and PPTIDs (2/2), and female preponderance in pineocytomas (3/3). Reis et al 4 reported 2 female cases of pineocytoma, and Komakula et al 18 reported male predilection in PPTIDs (7/11). There were equal numbers in both sexes for pineoblastoma in our series. For the PTPR, we reported 1 female and 1 male, in contrast with the previous study, 19 which reported 4 female cases. More cases are needed to assess sexual predilection.
The recent literature that suggests that the common pineal tumors have no pathognomonic imaging pattern 4 was supported quantitatively by our result that showed no significant difference in T1-and T2-signal intensity values between pineal parenchymal tumors and germinomas. The pineocytomas, PPTIDs, and pineoblastomas were reported as showing isohypointensity on T1WI and isohyperintensity on T2WI. 3,4,18 For the papillary tumors of the pineal region, variable signal intensity on T1WI and hyperintensity on T2WI can be seen. 2 Chang et al 19 and Vaghela et al 20 reported hyperintensity on T1WI that may be related to the concentration of the protein content in the cystic spaces. Most of the pineal germinomas in Wang et al 21 showed hypointensity on T1WI (19/22) and isointensity (26/32) on T2WI, whereas isointensity on both T1WI and T2WI was reported in 3 cases in Reis et al. 4 Our results showed that most of the pineal region tumors were categorized as showing heterogeneous enhancement, which correlated with other studies. 4,18,19 However, the groups showed no significant difference in coefficient of variation in T1WI, T2WI, and ADC images, which suggested that there was no significant difference in the degree of inhomogeneity.
All 20 patients in our study depicted variable degrees of enhancement, from mild to marked. The pineal parenchymal tumors showed nearly even distribution in degree of enhancement, while the previous studies reported strong enhancement in most of the PPTID cases, 18 mild enhancement in papillary tumors, 19 and avid enhancement in pineoblastoma cases. 2 In contrast with our study, in which most of the germinomas were observed as mildly enhancing (57.1%), Wang et al 21 reported marked enhancement in most of the cases (28/32). This may be caused by the variations in the postcontrast delay in scan time. While we administer gadolinium and perform a T2-weighted sequence, and then do 2 to 3 postcontrast T1-weighted sequences, we graded the degree of enhancement on the immediate axial postgadolinium T1WI, not the subsequent coronal and sagittal scans.
We found a statistically significant higher mean ADC value in germinomas (mean ADC 1590.69 ϫ 10 Ϫ6 mm 2 /s Ϯ 532.96 ϫ 10 Ϫ6 mm 2 /s; Fig 2) compared with pineal parenchymal tumors (mean ADC 883.58 ϫ 10 Ϫ6 mm 2 /s Ϯ 317.48 ϫ 10 Ϫ6 mm 2 /s; P ϭ .002). Douglas-Akinwande et al 22 also reported 11 cases of germinoma in which the solid portions showed 55% normal diffusion (mean ADC 947.64 ϫ 10 Ϫ6 mm 2 /s), 36% decreased diffusion (mean ADC 694.71 ϫ 10 Ϫ6 mm 2 /s), and 9% increased diffusion (mean ADC 1172.30 ϫ 10 Ϫ6 mm 2 /s). Because our study placed the ROIs around the tumor border at the maximum diameter, which included both cystic and solid portions, it likely caused the higher ADC value than the previously mentioned study. The lower ADC value in pineal parenchymal tumors might be an effect from the low ADC value in pineoblastomas (ADC 375.63-1027.49 ϫ 10 Ϫ6 mm 2 /s, mean ADC 614.06 Ϯ 239.65 ϫ 10 Ϫ6 mm 2 /s) (Fig 3). Low ADC value in the solid part of pineoblastomas has also been reported (mean ADC 512.2 Ϯ 24.4 ϫ 10 Ϫ6 mm 2 /s). 23 Moreover, a negative correlation of the ADC value in pineal PNETs (medulloblastoma, PNET, pineoblastoma) and germ cell tumors was suggested. However, there was no significant difference in ADC values between these 2 in a previous report. 24 The low ADC value in pineoblastoma might be explained by the high cellularity and high nuclear/cytoplasmic ratio of the tumor. Wide ranges of ADC values in germinomas might be caused by the difference amount in the cystic component and less attenuated tumor cellularity and lesser nuclear/cytoplasmic ratio as compared with pineoblastomas.
Data regarding ADC values of the pineocytoma, PPTID, and PTPR have yet to be established. Inoue et al 25  By setting a threshold value of 1250.00 ϫ 10 Ϫ6 mm 2 /s for ADC, we were able to predict the correct diagnosis in 17 of 19 pineal region lesions, doing better than the clinical impression of the experienced neuroradiologist basing his opinion on other imaging characteristics (11/20). This value should be tested in follow-up studies to determine if it remains valid across other study samples.
The small number of patients limited our retrospective study; therefore, a larger study to evaluate applicability of the 1250.00 ϫ 10 Ϫ6 mm 2 /s ADC for obtaining a differential diag-nosis of germinoma, pineocytoma, PPTID, PTPR, and pineoblastoma is suggested. Partial volume averaging at the tumor-CSF interface and from thick sections may lead to inaccurately high ADC. Difference in b values can interfere in the measurement: A recent study reported a statistically significant lower ADC value with a b-value of 1000 seconds/mm 2 as compared with a b-value of 800 seconds/mm 2 (P Ͻ .01). 26 Our study was also limited by the retrospective method, and some scan parameters could not be retrieved. Variable scanning parameters could have influenced our T1, T2, and ADC values. The assessment of enhancement was subjective and not quantitative. Also, the histopathology of tumors before 2007 must also be reviewed, because PPTID and PTPR were only recently recognized in the newer WHO classification in 2007. However, this would not affect our main intent in differentiating generically pineal cell versus germinoma tumors.

Conclusions
Germinomas showed higher ADC values than the pineal cell tumors (P ϭ .02), and the patients were younger. A threshold ADC value of 1250.00 ϫ 10 Ϫ6 mm 2 /s encircling the entire tumor was useful in distinguishing between germinomas and pineal cell origin tumors, with 89.5% accuracy. Otherwise, there were no definitive imaging characteristics that distinguished pineal cell from germinoma pineal region tumors. The T1WI and T2WI intensity values, degree of heterogeneity, and enhancement did not differentiate the 2.