Mishra Arabinda, Lu Yonggang, Meng Jingjing, Anderson Adam W, Ding Zhaohua
Institute of Imaging Science, Vanderbilt University, Nashville, 1161 21st Avenue South, MCN CCC-1118, TN 37232-2657, USA.
Neuroimage. 2006 Jul 15;31(4):1525-35. doi: 10.1016/j.neuroimage.2006.02.031. Epub 2006 Apr 19.
To enhance the performance of diffusion tensor imaging (DTI)-based fiber tractography, this study proposes a unified framework for anisotropic interpolation and smoothing of DTI data. The critical component of this framework is an anisotropic sigmoid interpolation kernel which is adaptively modulated by the local image intensity gradient profile. The adaptive modulation of the sigmoid kernel permits image smoothing in homogeneous regions and meanwhile guarantees preservation of structural boundaries. The unified scheme thus allows piece-wise smooth, continuous and boundary preservation interpolation of DTI data, so that smooth fiber tracts can be tracked in a continuous manner and confined within the boundaries of the targeted structure. The new interpolation method is compared with conventional interpolation methods on the basis of fiber tracking from synthetic and in vivo DTI data, which demonstrates the effectiveness of this unified framework.
为提高基于扩散张量成像（DTI）的纤维束成像性能，本研究提出了一个用于DTI数据各向异性插值和平滑的统一框架。该框架的关键组件是一个各向异性Sigmoid插值核，它由局部图像强度梯度轮廓自适应调制。Sigmoid核的自适应调制允许在均匀区域进行图像平滑，同时保证结构边界的保留。因此，该统一方案允许对DTI数据进行分段平滑、连续和边界保留插值，从而可以以连续方式跟踪平滑的纤维束并将其限制在目标结构的边界内。基于从合成和体内DTI数据进行的纤维跟踪，将新的插值方法与传统插值方法进行了比较，这证明了该统一框架的有效性。
