Adaptive spatial smoothing of fMRI images

Abstract

It is common practice to spatially smooth fMRI dataprior to statistical analysis and a number of differentsmoothing techniques have been proposed (e.g., Gaussiankernel filters, wavelets, and prolate spheroidal wave func-tions). A common theme in all these methods is that theextent of smoothing is chosen independently of the data,and is assumed to be equal across the image. This can leadto problems, as the size and shape of activated regions mayvary across the brain, leading to situations where certain re-gions are under-smoothed, while others are over-smoothed.This paper introduces a novel approach towards spatiallysmoothing fMRI data based on the use of nonstationaryspatial Gaussian Markov random fields (Yue and Speckman,2009). Our method not only allows the amount of smooth-ing to vary across the brain depending on the spatial ex-tent of activation, but also enables researchers to study howthe extent of activation changes over time. The benefit ofthe suggested approach is demonstrated by a series of sim-ulation studies and through an application to experimentaldata.Keywords and phrases: Spatially adaptive smooth-ing, Temporally adaptive smoothing, fMRI, Brain imaging,Smoothing.