Registration of contours of brain structures through a heat-kernel representation of shape
Jonathan R. Bates, Ying Wang, Xiuwen Liu, Washington Mio · 2009
We develop an algorithm for the registration of surfaces representing the contours of various subcortical structures of the human brain. We employ a scale-space representation of shape based on the heat kernel, which only depends on the intrinsic geometry of the surfaces. The multi-scale representation is used in conjunction with the non-linear Iterative Closest Point algorithm based on thin-plate-spline warps to establish point correspondences between shapes. The method is applied to the registration of the contours of four subcortical structures: the hippocampus, caudate nucleus, putamen, and third ventricle.