Three-dimensional linear and nonlinear transformations: an integration of light microscopical and MRI data.

The registration of image volumes derived from different imaging modalities such as MRI, PET, SPECT, and CT has been described in numerous studies in which functional and morphological data are combined on the basis of macrostructural information. However, the exact topography of architectural details is defined by microstructural information derived from histological sections. Therefore, a technique is developed for integrating micro- and macrostructural information based on 1) a three-dimensional reconstruction of the histological volume which accounts for linear and nonlinear histological deformations, and 2) a two-step procedure which transforms these volumes to a reference coordinate system. The two-step procedure uses an extended principal axes transformation (PAT) generalized to affine transformations and a fast, automated full-multigrid method (FMG) for determining high-dimensional three-dimensional nonlinear deformations in order to account for differences in the morphology of individuals. With this technique, it is possible to define upwards of 1,000 times the resolution of approximately 1 mm in MRI, making possible the identification of geometric and texture features of microscopically defined brain structures.