基于全变差空间正则化的纤维方向分布估计

发布时间：2018-11-14 16:08

【摘要】：提高白质纤维交叉重构能力是有效提高纤维跟踪技术的前提之一,目前大多纤维重构方法都是基于白质体素的独立重构,没有考虑到纤维的连续性特征,这就促使文章从全局范围考虑提高白质纤维重构能力.文章提出了一种基于全变差空间正则化的纤维方向分布估计方法,该方法首先利用字典基分布的球面反卷积策略拟合多壳采样信号,为了能够适用于单壳和多壳采样方案,文章重新定义了广义的纤维响应函数;进而在q空间中定义基函数系数的全变差正则化约束,旨在减少不必要的方向信息,降低因噪声引起的方向偏差,以获得纤维方向的空间局部一致性.实验分别在模拟数据和实际数据下进行,分别采用单壳和多壳数据验证了文章所提方法能够以更高效的性能实现纤维方向估计,相对于其他算法显著提高了纤维的连续性.
[Abstract]:Improving the ability of white matter fiber cross-remodeling is one of the prerequisites to improve the fiber tracking technology. At present, most of the methods of fiber reconstruction are based on the independent reconstruction of white matter voxel, and do not take into account the continuity characteristics of fibers. This prompted the article to improve the ability of white matter fiber remodeling from the global scope. In this paper, a method for estimating the direction distribution of fibers based on total variation space regularization is proposed. Firstly, the spherical deconvolution strategy of dictionary base distribution is used to fit the multi-shell sampling signals, in order to be suitable for single-shell and multi-shell sampling schemes. The generalized fiber response function is redefined in this paper. Then the regularization constraint of the base function coefficients is defined in QZone, which aims to reduce the unnecessary direction information and the direction deviation caused by noise, so as to obtain the spatial local consistency of fiber direction. The experiments were carried out under the simulated data and the actual data, respectively. The single-hull and multi-shell data were used to verify that the proposed method can realize the fiber direction estimation with more efficient performance. Compared with other algorithms, the fiber continuity is improved significantly.
【作者单位】：浙江工业大学信息工程学院;
【基金】：国家自然科学基金(61379020,61075062) 浙江省自然科学基金(LY13F0300067) 温州医科大学重中之重学科基金(LKFJ014)资助课题
【分类号】：TP391.41

【相似文献】