Purpose: Sometimes, cerebral lymphomas are difficult to distinguish from glioblastomas based on routine magnetic resonance (MR) examination. Thus far, the apparent diffusion coefficient (ADC) and different patterns of contrast enhancement have been valuable for the differential diagnosis of cerebral lymphomas and glioblastomas^{1, 2}. The ADC is based on a quantitative value; however, contrast-enhanced T1-weighted images (CET1WI) are not based on a quantitative value, but on appearance. It remains unclear whether there is an objective index available from the distribution of the intratumoral signal intensity on CET1WI, because no suitable statistical evaluation is available to date. Therefore, this study evaluated the utility of a histogram of the intratumoral enhanced region, using CET1WI, of cerebral lymphomas and glioblastomas, to determine whether the statistical parameters of the histogram (average, standard deviation, kurtosis, skewness) differ between the two tumors.

Materials and Methods: CET1WI were acquired in 20 patients with cerebral lymphomas (n = 10) or glioblastomas (n = 10) using a 3D-spoiled gradient recalled aquisition in the steady state (SPGR) sequence. All tumors were pathologically confirmed. MR imaging was performed using a 3.0-T whole body scanner (Discovery MR750, GE Healthcare, Waukesha, WI, USA) and 32-channel head coil. The patients were examined using CET1WI. Imaging parameters were as follows: FOV = 240 mm², matrix size = 420 × 192, slice thickness = 1.4 mm, TR/TE/TI = 7.2/2.2/700 ms, number of slices = 192 to 270. A standard dose of contrast agent (0.1 mmoL/kg) was injected. To reveal the contrast-enhanced area in the tumor, the region surrounding the outermost layer of the tumor was removed three-dimensionally on a workstation (ZIOSTATION; Amin Co., Ltd., Tokyo, Japan). The distribution of the signal intensity was measured and the average and standard deviation of the distribution of the signal intensity were automatically obtained. Histograms of the signal intensity within the tumor were generated in Excel. Differences between cerebral lymphomas and glioblastomas were examined using Mann-Whitney U tests.

Results: There was no significant difference in kurtosis between lymphomas (range: 1.91 to 3.75, median: 2.57) and glioblastomas (range: 0.17 to 2.91, median: 2.31). Skewness differed significantly (p < 0.01) between lymphomas (range: -0.94 to -0.09, median: -0.33) and glioblastomas (range: -0.30 to 0.38, median: 0.06; Fig. 1). There were no significant differences in the average and standard deviation between the two tumor types. We show representative cases of lymphoma and glioblastoma in Figs. 2 and 3, respectively.

Discussion: This study demonstrated the skewness of the signal intensity histogram of CET1WI significantly differed between cerebral lymphoma and glioblastoma. Even if the signal intensity per se of the MRI identifies the region of interest in CET1WI in each patient, it does not provide absolute values that are comparable among patients. In contrast, the skewness of the distribution of signal intensity values in CET1WI can be compared among patients. Unlike the conventional method of visually assessing contrast-enhancement area of the tumor in CET1WI, the distribution-based method can evaluate intratumoral signal distributions in CET1WI by an objective metric. Consequently, we found quantitative differences between lymphoma (homogeneous enhancement effect) and glioblastoma (heterogeneous enhancement effect); the skewness metric reflected a difference between the internal structure of lymphoma and glioblastoma. This suggests the possibility of differential diagnosis of cerebral lymphomas and glioblastomas without the need for special MR sequences.

Conclusion: The skewness calculated by the signal intensity histogram of CET1WI is a useful metrics for differential diagnosis between cerebral lymphomas and glioblastomas.