Savadjiev Peter, Campbell Jennifer S W, Descoteaux Maxime, Deriche Rachid, Pike G Bruce, Siddiqi Kaleem
Centre for Intelligent Machines and School of Computer Science, McGill University, Montréal, Canada.
Neuroimage. 2008 May 15;41(1):58-68. doi: 10.1016/j.neuroimage.2008.01.028. Epub 2008 Feb 1.
Whereas high angular resolution reconstruction methods for diffusion MRI can estimate multiple dominant fibre orientations within a single imaging voxel, they are fundamentally limited in certain cases of complex subvoxel fibre structures, resulting in ambiguous local orientation distribution functions. In this article we address the important problem of disambiguating such complex subvoxel fibre tract configurations, with the purpose of improving the performance of fibre tractography. We do so by extending a curve inference method to distinguish between the cases of curving and fanning fibre bundles using differential geometric estimates in a local neighbourhood. The key benefit of this method is the inference of curves, instead of only fibre orientations, to model the underlying fibre bundles. This in turn allows distinct fibre geometries that contain nearly identical sets of fibre orientations at a voxel, to be distinguished from one another. Experimental results demonstrate the ability of the method to successfully label voxels into one of the above categories and improve the performance of a fibre-tracking algorithm.
虽然扩散磁共振成像的高角分辨率重建方法可以估计单个成像体素内的多个主要纤维方向，但在某些复杂的亚体素纤维结构情况下，它们从根本上受到限制，导致局部方向分布函数不明确。在本文中，我们解决了消除此类复杂亚体素纤维束配置歧义的重要问题，目的是提高纤维束成像的性能。我们通过扩展一种曲线推理方法来实现这一点，该方法使用局部邻域中的微分几何估计来区分弯曲纤维束和扇形纤维束的情况。该方法的关键优势在于推断曲线，而不仅仅是纤维方向，以对潜在的纤维束进行建模。这反过来又允许区分在体素处包含几乎相同纤维方向集的不同纤维几何形状。实验结果证明了该方法能够成功地将体素标记为上述类别之一，并提高纤维跟踪算法的性能。
