Apparent diffusion coefficient (ADC) is a composite measure of Gaussian diffusion and perfusion. Several non-Gaussian diffusion imaging methods have been proposed to further estimate diffusion heterogeneity ^1,2 and perfusion parameters (e.g., cerebral blood volume, CBV) ³. Cellularity and vascularity are two critical indexes in brain tumor staging/diagnosis. Correlation has been reported between ADC and cellularity ⁴ and between CBV and vascularity ⁵. A recent study ⁶ combined diffusion kurtosis (DK) imaging and intravoxel incoherent motion (IVIM) imaging for simultaneous contrast-free measurement of intravascular fraction (f, a surrogate measure of CBV), diffusion coefficient (D, a measure of diffusivity), and diffusion kurtosis coefficient (K, a measure of diffusion heterogeneity):

$$\frac{S(b)}{S_{0}}\approx(1-f)exp(-bD+\frac{1}{6}b^{2}D^{2}K) $$

where S(b) and S₀ are the signal obtained with and without diffusion weighting (quantified by b-value b), respectively. The method was shown robust by simulations and in healthy volunteers, but has not been validated in clinical settings. In the present study, we investigated the feasibility of the hybrid method in glioma assessment.

Twenty-five patients with histologically proven brain gliomas were prospectively recruited. Tumors that had been treated were included if they showed definitive signs of residual or recurrent tumor at conventional MR imaging. This study was approved by the internal review board. All participants provided written informed consent. MR imaging was performed on a 3T system, using the body coil transmitter and a 12-channel phased-array receiver. Imaging protocol included: fluid attenuated inversion recovery (FLAIR), T2-weighted imaging, diffusion imaging, pre- and post-contrast T1-weighted imaging. A single-shot twice-refocused spin-echo echo-planar sequence was used for diffusion imaging: TR = 3.8 s, TE = 94 ms, field-of-view = 20 cm, in-plane matrix = 98x98, generalized autocalibrating partially parallel acquisition acceleration factor = 3, 18 slices parallel to the anterior commissure-posterior commissure line, slice thickness = 4 mm, b = 0, 400, 600, 850, 1200, 1700 s/mm², 8 repetitions after 1 dummy scan. Diffusion encoding was applied along 3 orthogonal directions in separate series which together took ~10 min.

f, D, and K were estimated on a voxel-wise basis by fitting the above equation to diffusion data after motion correction. By referring to conventional images, two raters independently defined the following regions of interest: contrast-enhanced tumor (hyperintensity in the post-contrast T1-weighted image with necrosis excluded), peritumoral edema (hyperintensity in the FLAIR image), and normal-appearing white matter (absence of abnormal hyperintensity and hypointensity in any of the anatomic images). Difference was resolved by consensus.

The area-dependent difference in f, D, and K was examined with repeated-measures multivariate analysis of variance, followed by post-hoc analyses. The correlation between f, D, and K was assessed in terms of Pearson’s correlation coefficient (r) and tolerance. Diagnostic performance was assessed in terms of the area under the curve (AUC) derived from receiver operating characteristic analysis. A significant level of 0.05 was used.

Repeated-measures multivariate analysis of variance revealed significant difference among areas (peritumoral edema, contrast-enhanced tumor, and normal-appearing white matter) (Wilks' λ = 0.028, F(6, 19) = 108.109, p < 10^-3). Univariate analysis revealed area-dependence in all three indexes: Greenhouse-Geisser F(1.302, 31.252) = 59.753, p < 10^-3 for D, F(1.789, 42.927) = 108.291, p < 10^-3 for K, and F(1.654, 39.688) = 17.666, p < 10^-3 for f. Post hoc analysis further indicated that D is lowest in normal-appearing white matter and f is lowest in peritumoral edema. K is greatest in normal-appearing white matter, followed by contrast-enhanced tumor and peritumoral edema.

D and K are closer correlated with each other (r = -0.77) than they are with f (r = -0.37 and 0.35, respectively). Yet, f, D, and K exhibit low collinearity (tolerance = 0.41 – 0.87), suggesting that they provide complementary information to one another. The AUC for distinguishing contrast-enhanced tumor is greatest with K, followed by f and D (see Figures 1 and 2).

While tumor cell growth almost always precedes angiogenesis, the association between vascularity (which may be assessed by f) and cellularity (which may be assessed by D) may vary with tumor stages as well as grades. The varied association changes the complexity of local microenvironment, to which K is previously suggested sensitive. Together, f, D, and K may better demarcate brain gliomas through exploration of multiple pathophysiological aspects.