Quantitative Diffusion Tensor Analysis Using Multiple Tensor Ellipsoids Model and Tensor Field Interpolation at Fiber Crossing

Abstract

In regions of intravoxel fiber crossing, the single-tensor model does not provide accurate results. The previously published models could resolve this issue but needed a long scan time and long computational time. This article aims to present the new model, which uses interpolated diffusion tensor orientations and requires the estimation of fewer parameters than the previously published model, where all parameters for the two diffusion ellipsoids have to be estimated. Fiber orientation information was reconstructed by using the radial basis function-based interpolation technique from tensor information in given seed regions of interest. Synthetic phantom data were generated, and the proposed method was compared with the conventional two-ellipsoid method. Data from one normal volunteer were analyzed to determine the effectiveness of the proposed method. The number of parameters to be estimated could be reduced by using the estimated fiber orientation information so that diffusion parameter calculation at fiber crossing becomes robust. The human study showed that fractional anisotropy (FA) values estimated by the proposed method (FA = 0.67 for the corpus callosum, 0.65 for the corticospinal tract) were significantly higher than that estimated by the standard single-tensor-based method (FA = 0.35), and the estimated FA value showed good agreement with the FA value in the adjacent fiber bundle. The proposed radial basis function-based technique could reconstruct diffusion properties at the fiber-crossing volume from sparse sampling of high angular diffusion weighted images.