On matching brain volumes

To characterize the complex morphological variations that occur naturally in human neuroanatomy so that their confounding eeect can be minimized in the identiication of brain structures in medical images, a computational framework has evolved in which individual anatomies are mod-eled as warped versions of a canonical representation of the anatomy, known as an atlas. To realize this framework, the method of elastic matching was invented for determining the spatial mapping between a 3-D image pair in which one image volume is modeled as an elastic continuum that is deformed to match the appearance of the second volume. In this paper, we review the primary concepts underlying the elastic matching, consider the implications of an integral formulation of the problem on its solution, and explore a more general Bayesian interpretation of the method in order to address issues that are otherwise diicult or unnatural to resolve within a continuum mechanical setting, including the examination of a solution's reliability or the incorporation of empirical information that may be available about the spatial mappings into the analysis. 1 Introduction Relying on the knowledge that diierent instances of the same anatomy share a common topological scheme, Bajcsy and Broit proposed that the various problems of localizing, measuring, or visualizing anatomic structures in a computed tomographic study could be reduced to the single problem of determining the spatial mapping between the anatomy depicted in the study and a reference, or atlas, image of the same anatomy 1, 2]. Using this mapping, information encoded in the atlas|such as anatomic labels and structural boundaries|is transferred to the image study. The transformed atlas labels automatically localize their corresponding structures in the tomographic images. Visualization is achieved by deforming, according to the mapping, the structural boundaries contained in the atlas and then displaying