Parametric Response Surface Models for Analysis of Multi-site fMRI Data
Analyses of fMRI brain data are often based on statistical tests applied to each voxel or use summary statistics within a region of interest (such as mean or peak activation). These approaches do not explicitly take into account spatial patterns in the activation signal. In this paper, we develop a response surface model with parameters that directly describe the spatial shapes of activation patterns. We present a stochastic search algorithm for parameter estimation. We apply our method to data from a multi-site fMRI study, and show how the estimated parameters can be used to analyze different sources of variability in image generation, both qualitatively and quantitatively, based on spatial activation patterns.
KeywordsPosterior Distribution Gibbs Sampler Precentral Gyrus Variance Component Analysis Conditional Posterior Distribution
- 2.Cao, J., Worsley, K.: Applications of random fields in human brain mapping. In: Moore, M. (ed.) Spatial Statistics: Methodological Aspects and Applications. Springer Lecture Notes in Statistics, vol. 159, pp. 169–182 (2001)Google Scholar
- 5.Hartvig, N.: A stochastic geometry model for fMRI data. Research Report 410, Department of Theoretical Statistics, University of Aarhus (1999)Google Scholar