Adaptive Smoothing as Inference Strategy
- PDF / 531,129 Bytes
- 11 Pages / 595.224 x 790.955 pts Page_size
- 41 Downloads / 213 Views
ORIGINAL ARTICLE
Adaptive Smoothing as Inference Strategy More Specificity for Unequally Sized or Neighbouring Regions Marijke Welvaert · Karsten Tabelow · Ruth Seurinck · Yves Rosseel
© Springer Science+Business Media New York 2013
Abstract Although spatial smoothing of fMRI data can serve multiple purposes, increasing the sensitivity of activation detection is probably its greatest benefit. However, this increased detection power comes with a loss of specificity when non-adaptive smoothing (i.e. the standard in most software packages) is used. Simulation studies and analysis of experimental data was performed using the R packages neuRosim and fmri. In these studies, we systematically investigated the effect of spatial smoothing on the power and number of false positives in two particular cases that are often encountered in fMRI research: (1) Single condition activation detection for regions that differ in size, and (2) multiple condition activation detection for neighbouring regions. Our results demonstrate that adaptive smoothing is superior in both cases because less false positives are introduced by the spatial smoothing process compared to standard Gaussian smoothing or FDR inference of unsmoothed data. Keywords False positive rate · fMRI · Gaussian smoothing · Power · Structural adaptive segmentation
M. Welvaert () · R. Seurinck · Y. Rosseel Department of Data Analysis, Ghent University, Ghent, Belgium e-mail: [email protected] K. Tabelow Weierstrass Institute for Applied Analysis and Stochastics, Berlin, Germany M. Welvaert Department of Data Analysis, H. Dunantlaan 1, 9000 Gent, Belgium
Introduction For many neuroscientists, spatial smoothing of fMRI data has become an automatic preprocessing step. The purpose of this smoothing procedure can be threefold. First, spatial smoothing moderates intersubject variation in brain anatomy, especially when individual brains are transformed to a standard brain space in order to allow intersubject comparison. Second, the smoothing of fMRI data will increase the signal-to-noise ratio (SNR). Third, a voxel-based massunivariate analysis of fMRI data calls for the need of multiple testing corrections. In this context, spatial smoothing enhances random field theory (RFT) based inference (see for example Worsley 2003). Although spatial smoothing, also referred to as spatial filtering, is mostly performed during the preprocessing stage of the analysis, it is actually a crucial step in the whole data analysis process because of its impact on the final result. In task-related fMRI studies, spatial smoothing will most likely result in an increase of the size of the activated region that is detected by the statistical analysis. The critical point here is that we need a useful estimate of the required width of smoothing. Depending on the goal of smoothing, guidelines vary substantially. To allow intersubject averaging, more smoothing might be necessary (e.g. a Gaussian kernel full-width half-maximum (FWHM) of 8 mm, which is often the default value in software package
Data Loading...
