Anatomically informed basis functions.

This paper introduces the general framework, concepts, and procedures of anatomically informed basis functions (AIBF), a new method for the analysis of functional magnetic resonance imaging (fMRI) data. In contradistinction to existing voxel-based univariate or multivariate methods the approach described here can incorporate various forms of prior anatomical knowledge to specify sophisticated spatiotemporal models for fMRI time-series. In particular, we focus on anatomical prior knowledge, based on reconstructed gray matter surfaces and assumptions about the location and spatial smoothness of the blood oxygenation level dependent (BOLD) effect. After reconstruction of the grey matter surface from an individual's high-resolution T1-weighted MRI, we specify a set of anatomically informed basis functions, fit the model parameters for a single time point, using a regularized solution, and finally make inferences about the estimated parameters over time. Significant effects, induced by the experimental paradigm, can then be visualized in the native voxel-space or on the reconstructed folded, inflated, or flattened cortical surface. As an example, we apply the approach to a fMRI study (finger opposition task) and compare the results to those of a voxel-based analysis as implemented in the Statistical Parametric Mapping package (SPM99). Additionally, we show, using simulated data, that the approach offers several desirable features particularly in terms of superresolution and localization.