Karkar Slim, Faisan Sylvain, Thoraval Laurent, Foucher Jack R
LSIIT/MIV (UMR 7005), CNRS, Université de Strasbourg.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:3497-500. doi: 10.1109/IEMBS.2009.5334578.
Most brain functional connectivity methods in fMRI require a brain parcellation into functionally homogeneous regions. In this work we propose a novel parcellation approach based on a spatial hierarchical clustering, that provides clusters within a multi-level framework. The method has the advantage of producing several brain parcellations rather than a single one from a fixed size-homogeneity criterion. Results obtained on real data demonstrate the relevance of the approach. Finally, a connectivity study shows the benefit of a prior multi-level parcellation of the brain.
功能磁共振成像（fMRI）中大多数脑功能连接方法都需要将大脑划分为功能均匀的区域。在这项工作中，我们提出了一种基于空间层次聚类的新型划分方法，该方法在多层次框架内提供聚类。该方法的优点是能从固定大小均匀性标准中产生多个脑部分割，而不是单个分割。在真实数据上获得的结果证明了该方法的相关性。最后，一项连接性研究显示了大脑先验多层次划分的益处。
