Role of PROPELLER diffusion weighted imaging and apparent diffusion coefficient in the diagnosis of sellar and parasellar lesions

Results ADC-MIN of hemorrhagic pituitary adenomas was lower than of the other lesions with similar appearance on conventional MRI (non-hemorrhagic pituitary adenomas, craniopharyngiomas, Rathke's cleft cysts; accuracy 100%); the useful cut-off value was 0.700 × 10 −3 mm 2 /s. ADC-MAX of meningiomas was lower than of non-hemorrhagic pituitary adenomas (accuracy 90.3%; p < 0.01). ADC-MIN of craniopharyngiomas was lower than of Rathke's cleft cysts (accuracy 100%; p < 0.05). Conclusion As PROPELLER DWI is less sensitive to susceptibility artifacts than single-shot echoplanar DWI, it is more useful in the examination of sellar and parasellar lesions. Calculation of the ADC values helps to differentiate between various sellar and parasellar lesions. Keywords Sella turcica Neoplasm Pituitary apoplexy Diffusion PROPELLER 1 Introduction Radiological imaging of the pituitary gland and parasellar region is challenging because the pituitary gland is a very small organ near many important structures. Magnetic resonance imaging (MRI) is the modality of choice; it provides multiplanar high-contrast images of the pituitary gland and adjacent structures [1] . Typical MRI protocols used for the evaluation of sellar masses include pre- and post-gadolinium-enhanced T1-weighted and T2-weighted coronal and sagittal sequences with a section thickness of 3 mm or less. However, it can be very difficult to distinguish between the sellar and parasellar tumors even with high-field MRI. DWI provides information on water mobility or diffusion within tissues by demonstrating Brownian motion in those tissues [2] and its role in the characterization and grading of brain tumors has been studied [3,4] . As echo-planar DWI, a widely accepted method, is often associated with strong susceptibility artifacts near the skull base, image distortion is a problem [2,5] . Periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) DWI was first described by Pipe et al. [6] . This sequence consists of a multishot-fast spin echo (FSE) sequence that takes advantage of both FSE methods with far fewer B 0 -related artifacts (from, for example metal, sinuses, and eddy currents) and PROPELLER MRI, which oversamples the region in the center of the k-space to correct for inconsistencies before combing the data [6] . This is the first report of the use of a high-field (3-T) MR imager and PROPELLER DWI to evaluate sellar and parasellar tumors. In this study we evaluated the benefit of adding PROPELLER DWI and ADC to conventional MRI for the diagnosis of these tumors. We also examined the value of calculating ADC-MIN, ADC-MEAN, and ADC-MAX for the differentiation among various sellar lesions. 2 Materials and methods We retrospectively studied sellar and parasellar lesions that were newly diagnosed and operated between October 2006 and February 2008 at Hiroshima University Hospital. During this period, 83 patients of sellar and parasellar lesions were operated in our institution but we excluded all tumors less than 10 mm in size ( n = 16) from this study to avoid partial volume effect on the measurement of regions of interest (ROI). Also, we excluded all clear Rathke's cleft cyst (RCC) from our study ( n = 7) because it could be easily differentiated from other sellar lesions on conventional MRI. Therefore, the study population consisted of 60 patients, 25 males and 35 females; their mean age was 49.5 ± 19.0 (range 5–93 years) and the mean size of the lesions was 25.35 ± 9.62 (range 11–51 mm). Of these, 24 had non-hemorrhagic and 14 had hemorrhagic pituitary adenoma, 10 had RCC, 7 had meningioma, and 5 had craniopharyngioma. Our study protocol was approved by the institutional review board of our institution; written patient consent was waived. To protect patient privacy we removed all identifiers from our records at the completion of our analyses. 2.1 MR sequences and image processing MR images were acquired on a 3-T superconducting system (Signa Horizon; GE Medical Systems, Milwaukee, WI). Coronal T1-WI was performed in the form of inversion recovery prepared fast spin-echo sequence with the following parameters: repetition time (TR) 2500 ms, echo time (TE) 16 ms, field of view (FOV) 18 cm × 18 cm, matrix size 256 × 192, 3-mm section thickness, 0.3 intersection gap, and one acquisition. The parameters for coronal FSE T2-WI were TR 4000 ms, TE 130 ms, FOV 18 cm × 18 cm, matrix size 256 × 192, 3-mm section thickness, 0.3 intersection gap, and one acquisition. For axial PROPELLER DWI the b -value was 1000, TR 6000 ms, TE 125 ms, FOV 22 cm × 22 cm, matrix size 128 × 128, 4-mm section thickness, and 0.3 intersection gap. After injecting 0.1 mmol/kg of a gadolinium-based contrast agent (gadopentate dimeglumine), coronal, sagittal, and axial T1-WI were acquired; the parameters were TR 2500 ms, TE 8 ms, FOV 18 cm × 18 cm, matrix size 256 × 192, 3-mm section thickness, 0.3 intersection gap, and 1 acquisition. We then transferred the MRI data to an off-line workstation (Advantage Workstation, AW 4.2, GE Medical Systems). The signal intensity (SI) of the lesions on conventional MRI sequences was measured by manually placing ROI in all parts of the lesions. Together, 2 authors (OM, FY, with 8 and 15 years of experience in brain MR imaging, respectively) who were blinded to the surgical and histological findings placed at least 6 ROI of uniform shape and size (elliptic, 50 mm 2 ) in the lesions; they avoided measurements in the clear cystic areas (areas which have similar MRI appearance to cerebrospinal fluid). The minimum, mean, and maximum SI on T1-WI and T2-WI, assessed by calculating the ratio of SI in the tumor to the SI of the normal subcortical frontal white matter in the same patient, was expressed as the minimum-, mean-, and maximum SI ratio (T1-WI-MIN, T1-WI-MEAN, T1-WI-MAX) and (T2-WI-MIN, T2-WI-MEAN, T2-WI-MAX), respectively. We also calculated the degree of tumor enhancement by dividing the SI in the sellar lesion on contrast enhanced T1-WI by the SI in the sellar lesion on non-enhanced T1-WI. The ADC was calculated with FuncTool software (GE Medical Systems). ADC maps were obtained by calculating the SI on DWI at 2 different b -values (0 and 1000) on a pixel-by-pixel basis. ADC values were measured by manually placing ROIs using the method applied for calculating SI on conventional MRI. The ADC values were expressed as minimum, mean, and maximum absolute values (ADC-MIN, ADC-MEAN, and ADC-MAX). 2.2 Operative data All 60 patients underwent complete resection. Intra-operatively, a neurosurgeon (AT, with 13 years of experience in pituitary surgery) determined the presence or absence of hemorrhage within the tumor. 2.3 Histological examination Surgical specimens were examined histologically by 2 neuropathologists (YT and VJA, with 11 and 10 years of experience in neuropathology) blinded to clinical and MRI data. 2.4 Statistical analysis For statistical analysis we used a commercially available software package (SPSS release 14.0 program, SPSS for windows; SPSS Inc., Chicago, IL). Logistic discriminant (regression) analysis was performed and the results were consensually determined by OM, FY, and MO to discriminate among various sellar lesions by using the SI ratios on T1- and T2-WI, the degree of enhancement, and absolute ADC values as independent variables. We compared the results between 2 groups of sellar lesions as follows: hemorrhagic vs. non-hemorrhagic pituitary adenomas, hemorrhagic pituitary adenomas vs. RCC, hemorrhagic pituitary adenomas vs. craniopharyngiomas, non-hemorrhagic pituitary adenomas vs. meningiomas, and craniopharyngiomas vs. RCC. In cases where the misclassification rate was zero (complete separation), maximum likelihood estimators and the significance of the estimated coefficients could not be obtained [7] . 3 Results All tumors in the sellar and parasellar region were clearly visualized on PROPELLER DWI; there was no image degradation. Conventional MRI findings on the sellar and parasellar lesions are listed in Table 1 ; ADC-MIN, ADC-MEAN, and ADC-MAX values are shown in Table 2 . 3.1 Non-hemorrhagic pituitary adenomas The solid portion of non-hemorrhagic pituitary adenomas ( n = 24) was hypointense to isointense on T1-WI and isointense to hyperintense on T2-WI. On contrast-enhanced images, it was homogeneously ( n = 11) and heterogeneously enhanced ( n = 13). On PROPELLER DWI, it appeared hyperintense ( n = 9), isointense ( n = 8), hypointense ( n = 2), and as mixed intensity ( n = 5). ADC-MIN, ADC-MEAN, and ADC-MAX values were 0.953 ± 0.181 (range 0.717–1.37), 1.08 ± 0.190 (range 0.787–1.50), and 1.20 ± 0.205 (range 0.911–1.67) × 10 −3 mm 2 /s, respectively ( Fig. 1 ). 3.2 Hemorrhagic pituitary adenomas Our study included 14 hemorrhagic pituitary adenomas. On T1-WI the tumors exhibited mixed intensity ( n = 9), isointensity ( n = 3), and hyperintensity ( n = 2). On T2-WI they exhibited mixed intensity ( n = 10), hypointensity ( n = 2), and hyperintensity ( n = 2). On contrast enhanced images, we found heterogeneous ( n = 11), homogeneous ( n = 2), and no enhancement ( n = 1). On PROPELLER DWI, hemorrhagic pituitary adenomas exhibited mixed intensity ( n = 10) and hyperintensity ( n = 4). As the hemorrhagic areas of the tumors showed remarkably varied SI, we measured the ADC in all parts of the tumors with the exception of clear cystic areas. ADC-MIN, ADC-MEAN, and ADC-MAX were 0.372 ± 0.162 (range 0.104–0.628), 0.711 ± 0.121 (range 0.419–0.977), and 1.03 ± 0.272 (range 0.606–1.61) × 10 −3 mm 2 /s, respectively ( Fig. 2 ). 3.3 Meningioma Our study included 7 meningiomas (tuberculum sellae, n = 4; paraclinoid meningiomas, n = 3). Histologically, 4 were meningothelial; one each was fibrous, transitional, and atypical. The tumors were hypointense to isointense on T1- and hyperintense on T2-WI; they showed homogenous gadolinium enhancement. On PROPELLER DWI, all were hyperintense. The ADC-MIN, ADC-MEAN, and ADC-MAX values were 0.768 ± 0.082 (range 0.686–0.894), 0.835 ± 0.095 (range 0.737–0.973), and 0.928 ± 0.166 (range 0.777–1.25) × 10 −3 mm 2 /s, respectively ( Fig. 3 ). 3.4 Rathke's cleft cysts (RCC) Our study included 10 symptomatic RCC; intraoperatively the cyst content was found to be mucinous. On T1-WI they appeared hypointense ( n = 2), isointense ( n = 2), and hyperintense ( n = 6). On T2-WI they were hyperintense ( n = 9) and hypointense ( n = 1). An intracystic waxy nodule exhibiting a hypointense signal on T2-WI was found in 5 cases; there was rim enhancement in 7 and no contrast enhancement in 3 cases. On PROPELLER DWI, RCC were hypointense ( n = 9) and hyperintense ( n = 1). The ADC-MIN, ADC-MEAN, and ADC-MAX values were 1.94 ± 0.448 (range 1.05–2.53), 2.02 ± 0.449 (range 1.11–2.60), and 2.13 ± 0.451 (range 1.17–2.64) × 10 −3 mm 2 /s, respectively ( Fig. 4 ). 3.5 Craniopharyngioma The 5 craniopharyngiomas were cystic and adamantinomatous. They showed mixed intensity on T1-and T2-WI ( n = 4), hyperintensity on T1- and mixed intensity on T2-WI ( n = 1). All craniopharyngiomas demonstrated rim enhancement. On PROPELLER DWI they appeared as mixed-intensity ( n = 4) and hyperintensity lesions ( n = 1). The ADC-MIN, ADC-MEAN, and ADC-MAX values were 1.38 ± 0.487 (range 0.749–2.02), 1.97 ± 0.535 (range 1.08–2.36), and 2.46 ± 0.602 (range 1.39–2.76) × 10 −3 mm 2 /s, respectively ( Fig. 5 ). 3.6 Differentiation among sellar lesions By logistic discriminant analysis, the accuracy of conventional MRI was 73.7% for the differentiation between hemorrhagic and non-hemorrhagic pituitary adenomas. It was 96% for the differentiation between hemorrhagic pituitary adenomas and RCC, and 94.7% for the differentiation between hemorrhagic pituitary adenomas and craniopharyngiomas. After adding the ADC values to conventional MRI findings for logistic discriminant analysis, we found that ADC-MIN was the most powerful variable for differentiating between hemorrhagic pituitary adenomas and other lesions with a similar MRI appearance. The ADC-MIN of hemorrhagic was lower than of non-hemorrhagic pituitary adenomas, craniopharyngiomas, and RCC with complete separation (accuracy 100%). We suggest that an ADC-MIN value of 0.700 × 10 −3 mm 2 /s is useful as a cut-off to differentiate between hemorrhagic pituitary adenomas and the other lesions with similar appearance on conventional MRI (non-hemorrhagic pituitary adenomas, craniopharyngiomas, and RCC). On conventional MRI we did not find a significant difference in signal intensity between non-hemorrhagic pituitary adenomas, meningiomas, and craniopharyngiomas and RCC. The addition of ADC values for logistic discriminant analysis revealed that ADC-MAX was the most powerful variable to differentiate between non-hemorrhagic pituitary adenomas and meningiomas (accuracy 90.3%, p < 0.01); ADC-MIN was the most powerful variable in the differentiation between craniopharyngiomas and RCC (accuracy 100%, p < 0.05). 4 Discussion There are few reports on the use of DWI for the diagnosis of sellar lesions [8,9] . Ours is the first study to employed PROPELLER DWI and 3.0 T-MRI to study sellar and parasellar lesions. On PROPELLER DWI the effect of susceptibility artifacts, common on single-shot echoplanar DWI especially in the sellar region, is decreased. We also took advantage of the higher signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and spatial resolution on 3.0 T-MRI compared to 1.5 T-MRI. In this study, we evaluated the benefits derived from adding ADC-MIN, ADC-MEAN, or ADC-MAX values to conventional MRI for the differential diagnosis of sellar lesions. The ADC-MIN was the most useful variable for differentiating between hemorrhagic pituitary adenomas and other lesions with a similar appearance on conventional MRI and between craniopharyngiomas and RCC. On the other hand, ADC-MAX was the most useful variable for differentiating between non-hemorrhagic pituitary adenomas and meningiomas. The ADC-MIN of hemorrhagic was significantly lower than of non-hemorrhagic pituitary adenomas with complete separation (accuracy 100%). We suggest that an ADC-MIN of 0.700 × 10 −3 mm 2 /s is useful as a cut-off value to differentiate between hemorrhagic and non-hemorrhagic pituitary adenomas. The lower ADC values of hemorrhagic pituitary adenomas can be explained by the susceptibility effect of blood products or by diffusion restriction in the presence of intact red blood cell membranes [10,11] . On conventional MRI, hemorrhage exhibits a wide range of appearances depending on its stage; this may decrease the accuracy of conventional MRI sequences in detecting hemorrhagic areas within pituitary adenomas; in our study the accuracy was 73.3%. The differentiation between hemorrhagic and non-hemorrhagic pituitary adenomas is important because hemorrhage may result in tumor enlargement and symptom aggravation [12] . Also, the presence of a hemorrhagic cavity often facilitates the removal of a macroadenoma, and such information is useful for surgical planning [13] . As the clinical presentation of hemorrhagic pituitary adenoma may not include pituitary apoplexy syndrome [14,15] , MRI is considered the diagnostic modality of choice to detect hemorrhage in pituitary adenomas. Tosaka et al. [15] who used T2*-weighted gradient-echo imaging to assess pituitary hemorrhage found that the presence of marked susceptibility artifacts in the sellar region limited the value of this method; applicability of T2*-weighted gradient echo imaging required suprasellar tumor extension and the exclusion of the anterior and inferior tumor portions from assessment. We found that the ADC-MIN value of hemorrhagic pituitary adenomas was significantly lower than of RCC with complete separation (accuracy 100%). This coincides with the finding of Kunii et al. [9] who used mean and relative ADC values derived from single-shot fast-spin echo DWI to differentiate between RCC and other sellar lesions in the pituitary fossa. Binning et al. [16] reported that the clinical and radiological differentiation between RCC and pituitary apoplexy is often difficult. They found that some RCC presented with acute-onset headache mimicking pituitary apoplexy. As the MRI appearance of hemorrhagic pituitary adenomas and RCC on MRI depends on the stage of hemorrhage and the RCC content, these lesions exhibit a wide, potentially identical range of SI. In addition, intracystic nodules, a common MRI finding in RCC cannot be distinguished easily from hemorrhage because both can exhibit a hyperintense signal on T1-WI and a hypointense signal on T2-WI [16] . Our observation that the ADC-MIN value of hemorrhagic pituitary adenomas was significantly lower than of craniopharyngiomas (accuracy 100%) may facilitate the differentiation between these lesions, which can be difficult on conventional MRI. Due to high protein and cholesterol concentrations, craniopharyngiomas usually show a hyperintense signal on T1- and a hyperintense or hypointense signal on T2-WI [17] . We found that the ADC of parasellar meningiomas was lower than of non-hemorrhagic pituitary adenomas. ADC-MAX was the most powerful variable for the differentiation between parasellar meningiomas and non-hemorrhagic pituitary adenomas (accuracy 90.3%, p < 0.01); there was no significant difference in the signal intensity between these lesions. We found that ADC-MAX was better than ADC-MIN and ADC-MEAN, possibly due to the wider range of ADC values in non-hemorrhagic pituitary adenomas that tend to be non-homogeneous compared to homogeneous meningiomas. Ours is the first study to describe the difference between the ADC of meningiomas and pituitary adenomas. The ADC values of our parasellar meningiomas were similar to the values reported for meningiomas using the same b- value [3,18] . Yamasaki et al. [3] detected no significant difference between the ADC value of parasellar meningiomas and pituitary adenomas, possibly because susceptibility artifacts on echoplanar DWI affected the ADC values in this region. The conventional MRI features of sellar and parasellar meningioma that distinguish them from pituitary macroadenomas are: (1) bright homogeneous gadolinium-enhancement compared to the heterogeneous relatively poor enhancement of pituitary adenomas, (2) a suprasellar rather than sellar tumor epicenter, and (3) tapered tumor extension into the intracranial dural base [1,19] . Meningiomas invading the cavernous sinus tend to constrict the carotid lumen; this is not a typical feature of pituitary adenomas [20] . As each of these findings can be subtle, a diagnosis based on conventional MRI sequences is sometimes difficult [19] . The ADC-MIN of craniopharyngiomas was significantly lower than of RCC (accuracy 100%, p < 0.05) and there was no significant difference in the signal intensity between these lesions. The lower ADC of craniopharyngiomas can be explained by the presence of keratin debris and higher cholesterol concentrations [9] . Craniopharyngiomas and RCC exhibit a wide SI range, this may render their differentiation difficult on conventional MRI. Although the ADC of craniopharyngiomas was lower than of RCC in our study, a few RCC manifesting a thick creamy content at surgery had low ADC values similar to those of craniopharyngiomas. This resulted in an overlap of ADC values between these lesions. Studies with larger numbers of cases are underway to assess the value of the ADC for differentiating between craniopharyngiomas and RCC. Our study is limited in that the conventional MRI findings included the signal intensity only without inclusion of other criteria for differentiating between sellar and parasellar lesions. However, previous MRI studies on sellar lesions indicated that these criteria can be subtle; therefore, in some cases the diagnosis remains controversial even if all criteria are applied. Based on our findings we conclude that PROPELLER DWI is less subject to susceptibility artifacts than single shot echoplanar DWI and that it is more useful for the examination of sellar and parasellar lesions. Calculation of ADC values is helpful in the differentiation among various sellar and parasellar lesions, especially hemorrhagic pituitary adenomas and non-hemorrhagic lesions with similar MRI appearance (non-hemorrhagic pituitary adenomas, RCC, and craniopharyngiomas). ADC was also useful for differentiating between pituitary adenomas vs. meningiomas and craniopharyngiomas vs. RCC. Conflict of interest None of the authors has to disclose any conflicts. Acknowledgements We thank Ursula Petralia for editorial review. This study was partially supported by the grant-in-aid for Young Scientists (Start-up), grants-in-aid from Japan Society for the Promotion of Science, the Ministry of Education, Culture, Sports, Science and Technology, Japan (grant No. 19890135). References [1] J. Rennert A. Doerfler Imaging of sellar and parasellar lesions Clin Neurol Neurosurg 109 2 2007 111 124 [2] R. Bammer Basic principles of diffusion-weighted imaging Eur J Radiol 45 3 2003 169 184 Review [3] F. Yamasaki K. Kurisu K. Satoh Apparent diffusion coefficient of human brain tumors at MR imaging Radiology 235 2005 985 991 [4] M. Provenzale S. Mukundan D.P. Barboriak Diffusion weighted and perfusion MR imaging for brain tumor characterization and assessment of treatment response Rev Radiol 239 3 2006 632 649 [5] D. Le Bihan C. Poupon A. Amadon F. Lethimonnier Artifacts and pitfalls in diffusion MRI J Magn Reson Imaging 24 3 2006 478 488 Review [6] J.G. Pipe V.G. Farthing K.P. Forbes Multishot diffusion-weighted FSE using PROPELLER MRI Magn Reson Med 47 1 2002 42 52 [7] A. Albert J.A. Anderson On the existence of maximum likelihood estimates in logistic regression models Biometrika 71 1984 1 10 [8] K. Okamoto J. Ito K. Ishikawa K. Sakai S. Tokiguchi Diffusion-weighted echo-planar MR imaging in differential diagnosis of brain tumors and tumor-like conditions Eur Radiol 10 2000 1342 1350 [9] K. Kunii T. Abe M. Kawamo D. Tanioka H. Izumiyama T. Moritani Rathke's cleft cysts: Differentiation from other cystic lesions in the pituitary fossa by use of single-shot fast spin-echo diffusion weighted MR imaging Acta Neurochir (Wien) 149 8 2007 759 769 [10] J.M. Rogg G.A. Tung G. Anderson S. Cortez Pituitary apoplexy: Early detection with diffusion weighted MR imaging AJNR Am J Neuroradiol 23 2002 1240 1245 [11] M. Wiesmann T.E. Mayer I. Yousry G.F. Hamann H. Brückmann Detection of hyperacute parenchymal hemorrhage of brain using echo-planar T2*-weighted and diffusion-weighted MRI Eur Radiol 11 2001 849 853 [12] N. Kurihara S. Takahashi S. Higano Hemorrhage in pituitary adenoma: Correlation of MR imaging with operative findings Eur Radiol 8 1998 971 976 [13] G. Mohr J. Hardy Hemorrhage, necrosis, and apoplexy in pituitary adenomas Surg Neurol 18 1982 181 189 [14] M. Piotin D. Tampieri D.A. Ruefenacht The various MRI patterns of pituitary apoplexy Eur Radiol 9 1999 918 923 [15] M. Tosaka N. Sato J. Hirato Assessment of hemorrhage in pituitary macroadenoma by T2*-weighted gradient-echo MR imaging AJNR Am J Neuroradiol 28 10 2007 2023 2029 [16] M.J. Binning J.K. Liu J. Gannon A.G. Osborn W.T. Couldwell Hemorrhagic and nonhemorrhagic Rathke's cleft cysts mimicking pituitary apoplexy J Neurosurg 108 1 2008 3 8 [17] F. Bonneville F. Cattin K. Marsot-Dupuch D. Dormont J.F. Bonneville J. Chiras T1 signal hyperintensity in the sellar region: Spectrum of findings Radiographics 26 1 2006 93 113 Review [18] U. Dorenbeck I.Q. Grunwald J. Schlaier S. Feuerbach Diffusion-weighted imaging with calculated apparent diffusion coefficient of enhancing extra-axial masses J Neuroimaging 15 4 2005 341 347 [19] S.L. Taylor J.A. Barakos G.R. Harsh IV C.B. Wilson Magnetic resonance imaging of tuberculum sellae meningiomas: preventing preoperative misdiagnosis as pituitary macroadenoma Neurosurgery 31 4 1992 621 627 [discussion 627] [20] J.L. Donovan G.M. Nesbit Distinction of masses involving the sella and suprasellar space: specificity of imaging features AJR Am J Roentgenol 167 1996 597 603