High order diffusion tensor imaging in human glioblastoma

Abstract

Rationale and objectives: Diffusion tensor imaging has been used to characterize tumor heterogeneity and invasion in human glioblastoma. Recently, higher order diffusion tensors have been proposed as solutions to errors associated with diffusion tensor imaging estimates of complex microstructures. The purpose of the current study was to examine higher order diffusion characteristics in human glioblastoma prior to surgical resection using the fourth-order diffusion tensor model.

Materials and methods: Twenty-five patients with newly diagnosed glioblastoma participated in the study. Diffusion-weighted images were collected in 21 directions. The second-order (traditional) and fourth-order diffusion tensors were calculated and compared in regions of contrast enhancement, T2 signal abnormality, and normal-appearing white matter.

Results: Orientation distribution functions were strikingly different between the two tensor models, particularly in regions with tumor heterogeneity and/or regions of suspected tumor invasion. Image contrast was significantly higher in fourth-order scalar measures compared to second-order scalars. Results of particular eigenvalues and scalars using the fourth-order tensor showed differences between T2 abnormal regions and contrast enhancement, whereas second-order eigenvalues and scalars did not show differences. This suggests that higher order diffusion images could potentially be more sensitive to tumor invasion.

Conclusions: These results suggest that the fourth-order diffusion tensor has the ability to add value to second-order (traditional) diffusion tensor imaging in the evaluation of glioblastoma.