Diffusion-based Heterogeneity Models in Magnetic Resonance Imaging for Characterization of Brain Tumors: An Introductory Study

Abstract

Recently developed diffusion-based magnetic resonance (MR) protocols have proven useful in assessing the heterogeneity of water diffusion in neural tissues, including brain tumors1. Based on theoretical increase in tumor cell heterogeneity compared to healthy brain2, these emerging imaging modalities offer several potentially useful applications, such as in identifying tumor margins, establishing tumor type and grade, and for differentiating tumor recurrence from post-treatment effect. In this study an introductory subset of five patients were scanned using a multi b-value Diffusion-Weighted Image (DWI) sequence, fitted with two previously described higher-order diffusion models. The first utilized a stretched exponential model (α-DWI)3; the second applied a cumulant expansion model (Diffusional Kurtosis Image, DKI)4. These models quantified water diffusion heterogeneity using the fitted parameters α and Kapp, respectively. The intent of this project was to gauge the potential utility of these MR models to apply diffusion heterogeneity information for characterization of brain tumors. Early results confirmed initial 6 hypotheses for high-grade gliomas, that (1) diffusion heterogeneity appeared greater in tumoral regions than in surrounding tissue, (2) high-grade tumors exhibited a relatively more heterogeneous diffusion pattern (lower α and higher K app) compared with low-grade glioma, and (3) the metastatic tumor had unique diffusion behavior compared to the primary tumors. Overall, this introductory study generally supports the potential ability of higher-order diffusion heterogeneity models to characterize brain tumors. More detailed investigation of this application across a larger subset of patients and tumor types may be beneficial.