Peyrat Jean-Marc, Delingette Hervé, Sermesant Maxime, Pennec Xavier, Xu Chenyang, Ayache Nicholas
INRIA - Asclepios Research Project, Sophia Antipolis, France.
Med Image Comput Comput Assist Interv. 2008;11(Pt 2):972-9. doi: 10.1007/978-3-540-85990-1_117.
In this paper, we propose a generic framework for intersubject non-linear registration of 4D time-series images. In this framework, spatio-temporal registration is defined by mapping trajectories of physical points as opposed to spatial registration that solely aims at mapping homologous points. First, we determine the trajectories we want to register in each sequence using a motion tracking algorithm based on the Diffeomorphic Demons algorithm. Then, we perform simultaneously pairwise registrations of corresponding time-points with the constraint to map the same physical points over time. We show this trajectory registration can be formulated as a multichannel registration of 3D images. We solve it using the Diffeomorphic Demons algorithm extended to vector-valued 3D images. This framework is applied to the inter-subject non-linear registration of 4D cardiac CT sequences.
在本文中,我们提出了一个用于4D时间序列图像的主体间非线性配准的通用框架。在这个框架中,时空配准是通过映射物理点的轨迹来定义的,这与仅旨在映射同源点的空间配准相反。首先,我们使用基于微分同胚恶魔算法的运动跟踪算法来确定每个序列中我们想要配准的轨迹。然后,我们对相应的时间点进行同时成对配准,并施加约束以随时间映射相同的物理点。我们表明,这种轨迹配准可以被表述为3D图像的多通道配准。我们使用扩展到向量值3D图像的微分同胚恶魔算法来解决它。这个框架被应用于4D心脏CT序列的主体间非线性配准。