Bezgin Gleb, Wanke Egon, Krumnack Antje, Kötter Rolf
Department of Cognitive Neuroscience, Section Neurophysiology and Neuroinformatics, Radboud University Medical Center, 6500HB Nijmegen, The Netherlands.
Neural Netw. 2008 Oct;21(8):1132-45. doi: 10.1016/j.neunet.2008.05.010. Epub 2008 Jun 11.
We propose a new technique, called Spatial Objective Relational Transformation (SORT), as an automated approach for derivation of logical relationships between cortical areas in different brain maps registered in the same Euclidean space. Recently, there have been large amounts of voxel-based three-dimensional structural and functional imaging data that provide us with coordinate-based information about the location of differently defined areas in the brain, whereas coordinate-independent, parcellation-based mapping is still commonly used in the majority of animal tracing and mapping studies. Because of the impact of voxel-based imaging methods and the need to attribute their features to coordinate-independent brain entities, this mapping becomes increasingly important. Our motivation here is not to make vague statements where more precise spatial statements would be better, but to find criteria for the identity (or other logical relationships) between areas that were delineated by different methods, in different individuals, or mapped to three-dimensional space using different deformation algorithms. The relevance of this problem becomes immediately obvious as one superimposes and compares different datasets in multimodal databases (e.g. CARET, http://brainmap.wustl.edu/caret), where voxel-based data are registered to surface nodes exploited by the procedure presented here. We describe the SORT algorithm and its implementation in the Java 2 programming language (http://java.sun.com/, which we make available for download. We give an example of practical use of our approach, and validate the SORT approach against a database of the coordinate-independent statements and inferences that have been deduced using alternative techniques.
我们提出了一种名为空间目标关系转换（SORT）的新技术，作为一种自动方法，用于推导在同一欧几里得空间中配准的不同脑图谱中皮质区域之间的逻辑关系。最近，出现了大量基于体素的三维结构和功能成像数据，这些数据为我们提供了关于大脑中不同定义区域位置的基于坐标的信息，而基于坐标独立的、基于脑区划分的图谱在大多数动物追踪和图谱研究中仍被普遍使用。由于基于体素的成像方法的影响以及将其特征归因于坐标独立的脑实体的需求，这种图谱变得越来越重要。我们的动机不是在更精确的空间陈述会更好的情况下做出模糊的陈述，而是要找到不同方法所描绘的、不同个体中的或使用不同变形算法映射到三维空间的区域之间的同一性（或其他逻辑关系）标准。当在多模态数据库（例如CARET，http://brainmap.wustl.edu/caret）中叠加和比较不同数据集时，这个问题的相关性立即变得明显，在该数据库中，基于体素的数据被配准到此处介绍的过程所利用的表面节点。我们描述了SORT算法及其在Java 2编程语言（http://java.sun.com/，我们提供该代码供下载）中的实现。我们给出了我们方法实际应用的一个例子，并针对使用替代技术推导的坐标独立陈述和推理的数据库验证了SORT方法。
