A factor analytic approach to structural characterization

The paper presents an exploratory and confirmatory factor analytic approach to morphometry in which a high-dimensional set of shape-related variables is examined with the purpose of finding clusters with strong col-relation. This clustering can potentially identify regions that have anatomic significance and thus lend insight to morphometric investigations. In the present work, shape is characterized in a new basis, in which the variables account for the correlation among structural regions of interest. The analysis is based on information about size differences between the differential volume about points in a template image and their corresponding volumes in a subject image, where the correspondence is established by non-rigid registration. The method is demonstrated in a study of sex differences in the human corpus callosum. We show that the regions where those differences occur can be deter-mined by unsupervised analysis. The most significant factors are used to represent the sample in a lower-dimensional basis, with which a prior model is constructed and used for classification.