Differentiation among glioblastoma multiforme, solitary metastatic tumor, and lymphoma using whole-tumor histogram analysis of the normalized cerebral blood volume in enhancing and perienhancing lesions

Abstract

Background and purpose: The histogram method has been shown to demonstrate heterogeneous morphologic features of tumor vascularity. This study aimed to determine whether whole-tumor histogram analysis of the normalized CBV for contrast-enhancing lesions and perienhancing lesions can differentiate among GBMs, SMTs, and lymphomas.

Materials and methods: Fifty-nine patients with histopathologically confirmed GBMs (n = 28), SMTs (n = 22), or lymphomas (n = 12) underwent conventional MR imaging and dynamic susceptibility contrast-enhanced imaging before surgery. Histogram distribution of the normalized CBV was obtained from whole-tumor voxels in contrast-enhancing lesions and perienhancing lesions. The HW, PHP, and MV were determined from histograms. One-way ANOVA was used initially to test the overall equality of mean values for each type of tumor. Subsequently, posttest multiple comparisons were performed.

Results: For whole-tumor histogram analyses for contrast-enhancing lesions, only PHP could differentiate among GBMs (4.79 ± 1.31), SMTs (3.32 ± 1.10), and lymphomas (2.08 ± 0.54). The parameters HW and MV were not significantly different between GBMs and SMTs, whereas the 2 histogram parameters were significantly higher in GBMs and SMTs compared with lymphomas. For the analyses of perienhancing lesions, only MV could differentiate among GBMs (1.90 ± 0.26), SMTs (0.80 ± 0.21), and lymphomas (1.27 ± 0.34). HW and PHP were not significantly different between SMTs and lymphomas.

Conclusions: Using a whole-tumor histogram analysis of normalized CBV for contrast-enhancing lesions and perienhancing lesions facilitates differentiation of GBMs, SMTs and lymphomas.

Fig 1.

The schema of histogram distribution in a patient with GBM. A and B, Axial postcontrast T1-weighted image (A) and the normalized CBV map (B) show typical findings of GBM. C, Three histogram parameters are defined as HW, PHP, and MV.

Fig 2.

MR images and dynamic susceptibility contrast-enhanced imaging histograms in patients with GBM (A), metastatic tumor (B), and lymphoma (C).

Fig 3.

Box-and-whisker plots of 6 histogram parameters for 3 pathologic groups of solitary enhancing brain lesions. A, HW in contrast-enhancing lesions. B, PHP in contrast-enhancing lesions. C, MV in contrast-enhancing lesions. D, HW in perienhancing lesions. E, PHP in perienhancing lesions. F, MV in perienhancing lesions.

Fig 4.

Comparison of ROC curves of 6 histogram parameters in the contrast-enhancing lesions and perienhancing lesions for differentiating GBM and SMT (A), GBM and lymphoma (B), and SMT and lymphoma (C).