基于局部特征的扩散张量场插值追踪方法

Interpolation of Diffusion Tensor Field Based on Local Features

基于扩散张量成像(diffusion tensor imaging,DTI)进行人脑白质神经纤维追踪可以实现活体无创重建纤维.由于DTI分辨率的局限性,在追踪过程中需要对扩散张量场进行插值以获取连贯的纤维成像.但是,新插入的扩散张量往往会受到奇异点附近张量的无方向性影响,而产生挤压效应与方向的不连续问题.提出一种基于局部特征的扩散张量场插值追踪方法,通过考虑扩散张量各向异性与距离影响来优化各扩散张量主方向的旋转方向,保持追踪过程中纤维的连续性;通过对特征值和特征向量分别插值,保持扩散张量的各向异性.最后,对2种典型的奇异点张量场进行插值,并与已有的典型的张量场插值方法相比较.实验结果表明,该方法有效地避免了扩散张量场中奇异点的干扰,成功地实现了对扩散张量场的纤维追踪.

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.