Spatial Smoothing with Robust Priors in Functional MRI
Functional magnetic resonance imaging (fMRI) has led to rapid advances in human brain mapping. The statistical assessment of brain areas that are activated by a particular stimulus requires elaborate models and efficient algorithms. Based on regressionmodels of the observedMR signals on the presented stimulus, pixelwise and spatial techniques have been applied. Besides describing the stimulus dependence of the MR signal, the latter also include spatial correlations between neighboring areas in the brain. In a Bayesian context mostly Gaussian Markov random field priors are used for this purpose, possibly blurring the smoothed images. To preserve potentially existing edges, this article presents some robust alternatives to the Gaussian choice. Inference is based on MCMC techniques. The performance of the new approaches will be demonstrated for simulated as well as for real data.
KeywordsFunctional Magnetic Resonance Imaging fMRI Data Markov Random Field Observation Model Spatial Smoothing
Unable to display preview. Download preview PDF.
- BESAG, J., YORK, J., and MOLLIE, A. (1991): Bayesian image restoration with two applications in spatial statistics. Annals of the Institute of Statistical Mathematics, 43–1, 1-59.Google Scholar
- GÖSSL, C. (2001): Bayesian Models in functional Magnetic Resonance Imaging: Approaches for Human Brain Mapping, Dissertation, Dept. of Statistics, Univ. of Munich, to appear in Shaker Verlag, Aachen.Google Scholar
- GÖSSL, C., AUER, D. P., and FAHRMEIR, L. (2001): Bayesian spatiotemporal inference in functional magnetic resonance imaging. Biometrics, In Press.Google Scholar
- GÖSSL, C., FAHRMEIR, L. and AUER, D. P. (2001): Bayesian Modeling of Hemodynamic Response Function in BOLD fMRI, NeuroImage, to appear.Google Scholar
- HENNERFEIND, A. (2000): Bayesianische kantenerhaltende Glättung, Diplomarbeit, Dept. of Statistics, Univ. of Munich.Google Scholar