To Smooth or Not to Smooth?: Bias and Efficiency in fMRI Time-Series Analysis

By Karl J. Friston, Oliver Josephs, E. Zarahn, A. Holmes, J B Poline. 2000

467 citations recorded by the source corpus.

Subfields: Cognitive Neuroscience, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/44a4b4907d7572c94ded13353d3dbeb898bb9101

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Multivariate Model Specification for fMRI Data (2002, methodology): “…see for instance Andersen et al., 1999; Baumgartner et al., 2000; Bullmore et al., 1996b; McKeown, 2000; Sychra et al., 1994; McIntosh et al., 1996; Friston et al., 1993, 1999, 2000b.) The next section gives an overview of multivariate analyses and explains our choice in the context of model…”
Two Distinct Neuronal Networks Mediate the Awareness of Environment and of Self (2011, methodology): “High-pass filtering using a cutoff period of 128 sec was implemented in the design matrix to remove low-frequency drift from the time series (Vanhaudenhuyse et al., 2009; Boly et al., 2007; Friston et al., 2000).”
Brain mediators of cardiovascular responses to social threat, Part I: Reciprocal dorsal and ventral sub-regions of the medial prefrontal cortex and heart-rate reactivity (2009, background): “We note that an autoregressive level-1 error structure was not specified in the model because although allowing for autocorrelation can produce maximally efficient inferential statistics, autocorrelation parameter estimates based on noisy level-1 data are not necessarily more efficient (Friston et a”
Colored noise and computational inference in neurophysiological (fMRI) time series analysis: Resampling methods in time and wavelet domains (2001, methodology): “, 1997; Purdon and Weisskoff, 1998; Dale, 1999]; or by predicting the off-diagonal elements of V by convolving the data and design matrix with a known smoothing kernel [Friston et al., 1995; Worsley and Friston, 1995; Friston et al., 2000].”
Statistical parametric mapping (2013, methodology): “Hemodynamics elicited by an impulse of neuronal activity as predicted by a dynamical biophysical model (see Friston et al 2000a for details ).”
Experimental Design and the Relative Sensitivity of BOLD and Perfusion fMRI (2002, background): “The removal of low frequencies and application of exogenous smoothing is necessary, even in the presence of the 1/f model, because there is substantial variability in the actual magnitude of the low-frequency power from voxel to voxel (Friston et al., 2000).”
Classical and Bayesian Inference in Neuroimaging: Applications (2002, background): “(A.7) in 3 As discussed in Friston et al. (2000) the GLS scheme was originally introduced to ensure robust estimates of variance in the face of unknown serial correlations.”
A General Statistical Analysis for fMRI Data (2000, unknown): “The removal of low frequencies implements an approximate whitening of the data and removal of high frequencies renders the resulting inference more robust to misspecification of the correlation structure (Friston et al., 2000).”
Temporal Autocorrelation in Univariate Linear Modeling of FMRI Data (2001, methodology): “Prewhitening requires a robust estimator of the autocorrelation to maintain low bias and Friston et al. (2000) suggest that current techniques for estimating the autocorrelation are not accurate enough to give prewhitening acceptable bias.”
A Developmental Examination of Amygdala Response to Facial Expressions (2008, methodology): “These normalized contrast images were then smoothed with an isotropic Gaussian kernel (full-width half-maximum = 11.4) to reduce nonstationarity in the spatial autocorrelation structure produced by the previous step (Friston et al., 2000).”
Estimating the number of independent components for functional magnetic resonance imaging data (2007, background): “It is also important to note that there is temporal dependence in both the fMRI signal and the noise, which have been studied by, e.g., Bullmore et al. (2001) and Friston et al. (2000) .”
WSPM: Wavelet-based statistical parametric mapping (2007, methodology): “For our framework, we choose to incorporate the method deployed by SPM (Worsley and Friston, 1995; Friston et al., 2000, 2002 ).”
Handbook of Functional MRI Data Analysis: Index (2011, background): “This was in part due to inaccurate correlation models that gave biased variance estimates and could inflate false positive risk (Friston et al., 2000).”
Multilevel linear modelling for FMRI group analysis using Bayesian inference (2004, background): “This inherently means that we assume that the autocorrelation is known with no uncertainty, an assumption that is commonly made in FMRI time series analysis (Bullmore et al., 1996; Friston et al., 2000; Woolrich et al., 2001).”
Support vector machine learning-based fMRI data group analysis (2007, methodology): “This is probably due to the limited number of subjects scanned and the lower SNR of ASL perfusion fMRI data compared to BOLD fMRI data.”