Interpolation of Diffusion Tensor Field Based on Local Features

Based on diffusion tensor imaging(DTI),tracking of human white matter nerve fibers can achieve non-invasive reconstruction of living fibers.Due to the limitations of DTI resolution,the diffusion tensor field needs to be interpolated during the tracking process to obtain continuous fiber imaging.However,the interpolated diffusion tensors are often affected by the non-directionality problem near the degenerate point(called isotropic tensor).This will cause the squeezing effect and the discontinuity of the direction.In this paper,we propose a local features-based interpolation method,which can successfully track the nerve fibers around the isotropic tensor.Our method optimizes the rotation direction of each tensor by considering the effect of tensor anisotropy and distance,it can maintain the continuity of the fibers during tracking;It also retains the anisotropy of the diffusion tensor by interpolating the eigenvalues and eigenvectors separately.Finally,in order to demonstrate the effectiveness of our proposed method,the two classical isotropic diffusion tensor fields are interpolated:Wedge and Trisector.Compared with existing interpolation methods,our method effectively reduced the interference of degenerate points,and successfully tracked the fibers in the two diffusion tensor fields.