Elsevier

NeuroImage

Volume 64, 1 January 2013, Pages 240-256
NeuroImage

An improved framework for confound regression and filtering for control of motion artifact in the preprocessing of resting-state functional connectivity data

https://doi.org/10.1016/j.neuroimage.2012.08.052Get rights and content

Abstract

Several recent reports in large, independent samples have demonstrated the influence of motion artifact on resting-state functional connectivity MRI (rsfc-MRI). Standard rsfc-MRI preprocessing typically includes regression of confounding signals and band-pass filtering. However, substantial heterogeneity exists in how these techniques are implemented across studies, and no prior study has examined the effect of differing approaches for the control of motion-induced artifacts. To better understand how in-scanner head motion affects rsfc-MRI data, we describe the spatial, temporal, and spectral characteristics of motion artifacts in a sample of 348 adolescents. Analyses utilize a novel approach for describing head motion on a voxelwise basis. Next, we systematically evaluate the efficacy of a range of confound regression and filtering techniques for the control of motion-induced artifacts. Results reveal that the effectiveness of preprocessing procedures on the control of motion is heterogeneous, and that improved preprocessing provides a substantial benefit beyond typical procedures. These results demonstrate that the effect of motion on rsfc-MRI can be substantially attenuated through improved preprocessing procedures, but not completely removed.

Highlights

► We describe spatial, temporal, and spectral features of rsfc-MRI motion artifact. ► We show how these artifact features impact preprocessing choices. ► We systematically evaluate different confound regression and filtering techniques. ► Our optimized preprocessing approach minimizes rsfc-MRI motion artifact.

Introduction

Although it has long been known that in-scanner head motion can have profound effects on fMRI timeseries data (Bullmore et al., 1999, Friston et al., 1996), the specific importance of this artifact for the analysis of resting state functional connectivity MRI (rsfc-MRI; Biswal et al., 1995, Fox and Raichle, 2007) has only recently been appreciated. In particular, it has been demonstrated in three large independent samples (Power et al., 2011a, Power et al., 2011b, Satterthwaite et al., 2012a, Van Dijk et al., 2011) that even relatively small amounts of in-scanner head motion represent a substantial confound for rsfc-MRI data. All three studies concluded that motion in general tends to enhance short-range connectivity and diminish long-distance connectivity among network nodes. As rsfc-MRI has evolved to become an important tool for examining brain networks in health and disease (Biswal et al., 2010, Fox and Greicius, 2010, Glahn et al., 2010, Seeley et al., 2009, Yeo et al., 2011, Zhou et al., 2010), it is of critical importance to understand how best to model and account for this artifact.

Power et al., 2011a, Power et al., 2011b recently introduced a novel method, called “scrubbing,” that identifies motion-induced spikes in the rsfc-MRI timeseries and excises these data with a temporal mask; adjacent timepoints are then temporally concatenated. Subsequently, Carp (2011) proposed a modification of scrubbing where data were removed and interpolated prior to band-pass filtering in order to avoid propagation of the motion artifact during the application of the band-pass filter. Using simulated data, he demonstrated that this modified scrubbing procedure was able to recover the “ground truth” connectivity in this timeseries (Carp, 2011). In a reply to Carp, Power et al. (2012) note that this procedure may be of marginal benefit given the fact that motion often occurs in long epochs, and that the effect of motion may occur beyond one isolated volume.

Scrubbing is a preprocessing technique that can be implemented after (Power et al., 2011a, Power et al., 2011b) or as part of (Power et al., 2012) standard rsfc-MRI preprocessing, which usually includes image re-alignment, spatial smoothing, filtering, and confound regression (Van Dijk et al., 2010). Notably, no prior report has investigated whether these standard rsfc-MRI preprocessing steps can themselves be improved to control the artifacts induced by in-scanner head motion. Here, we focus on two of these steps – confound regression and filtering – and investigate whether improved methods can produce better control of motion artifact.

There is substantial variation regarding how motion is modeled during confound regression (Auer, 2008, Johnstone et al., 2006): some studies include only the six motion parameters themselves, while others include the temporal derivatives, or even the quadratic of both the raw parameters and derivatives (zu Eulenburg et al., 2012). Yet other studies have modeled motion-induced spikes in the timeseries data with individual regressors, effectively removing the effect of these data points on any subsequent analysis of the residual timeseries (Lemieux et al., 2007). Furthermore, while most studies apply a band-pass filter with a high-pass cutoff in the range of 0.008–0.01 Hz and a low-pass threshold of 0.08–0.1 Hz (Cordes et al., 2001, Niazy et al., 2011), it has not yet been specifically demonstrated how motion affects the magnitude spectra of rsfc-MRI data, nor is it known whether band-pass filtering can be tailored for better control of motion artifact.

This study investigates the effect of motion and the improvement of preprocessing procedures in a large sample of adolescents (n = 348) who completed an rsfc-MRI study of typical duration (6 minutes). We had two primary goals. First, we sought to describe the spatial, temporal, and spectral characteristics of motion artifact, and evaluated how typical preprocessing steps alter the manifestations of this artifact. Second, we systematically evaluated whether confound regression and filtering could be improved to provide better control of motion artifact. Results reveal that the effectiveness of preprocessing procedures on the control of motion artifact are quite variable, and that improved preprocessing provides a substantial benefit beyond typical procedures, allowing the attenuation but not complete removal of motion artifact from rsfc-MRI data.

Section snippets

Overall approach

Reflecting the two main goals of this study, the methods and results of this paper are described in two parts. In the first section we further describe how in-scanner motion affects rsfc-MRI data through use of both real data and simulations, and how different preprocessing strategies may alter the way motion artifact manifests. In order to evaluate the spatial distribution of motion artifact, we introduce a novel procedure for estimating motion on a voxelwise basis. In the second section, we

Discussion

Motion artifact is a primary obstacle impeding the application of rsfc-MRI to the study of individual and group differences. Therefore, understanding and mitigating the influence of this artifact are critical. Our results further describe the spatial, temporal, and spectral characteristics of motion artifact in rsfc-MRI data, providing information that can guide efforts to minimize its effect. Systematic analyses indicate that improved confound regression and filtering can substantially

Disclosures

Authors report no disclosures.

Financial support

This study was supported by grants from the National Institute of Mental Health RC2MH089983 and RC2MH089924. Dr. Satterthwaite was supported by NIMH T32 MH019112, APIRE, and The Brain & Behavior Research Foundation through the Marc Rapport Family Investigator Grant. Dr. Wolf was supported by NIMH MH085096, APIRE, and The Brain & Behavior Research Foundation through the Sidney R. Baer, Jr. Foundation. Dr. Eickhoff was supported by the Human Brain Project (R01‐MH074457‐01A1) and the Helmholtz

Acknowledgments

Many thanks to the acquisition and recruitment team: Karthik Prabhakaran, Jeff Valdez, Marisa Riley, Rosetta Chiavacci, Ryan Hopson, Jack Keefe, and Nick DeLeo. James Dickson, Chad T. Jackson, Mark Griffin, and Tianyi Du provided data management and systems support. Thanks to Aaron Alexander Bloch for discussion of the permutation analysis, and to Jonathan Power, Mika Rubinov, and our anonymous reviewers for many helpful suggestions and discussion.

References (60)

  • R.K. Niazy et al.

    Spectral characteristics of resting state networks

    Prog. Brain Res.

    (2011)
  • J.D. Power et al.

    Functional network organization of the human brain

    Neuron

    (2011)
  • M. Rubinov et al.

    Weight-conserving characterization of complex functional brain networks

    Neuroimage

    (2011)
  • T.D. Satterthwaite et al.

    Impact of in-scanner head motion on multiple measures of functional connectivity: relevance for studies of neurodevelopment in youth

    Neuroimage

    (2012)
  • T.D. Satterthwaite et al.

    Being right is its own reward: load and performance related ventral striatum activation to correct responses during a working memory task in youth

    Neuroimage

    (2012)
  • V. Schöpf et al.

    Fully exploratory network ICA (FENICA) on resting-state fMRI data

    J. Neurosci. Methods

    (2010)
  • V. Schöpf et al.

    Model-free fMRI group analysis using FENICA

    Neuroimage

    (2011)
  • W.W. Seeley et al.

    Neurodegenerative diseases target large-scale human brain networks

    Neuron

    (2009)
  • K. Shmueli et al.

    Low-frequency fluctuations in the cardiac rate as a source of variance in the resting-state fMRI BOLD signal

    Neuroimage

    (2007)
  • S.C. Strother et al.

    The quantitative evaluation of functional neuroimaging experiments: the NPAIRS data analysis framework

    Neuroimage

    (2002)
  • J. Tohka et al.

    Automatic independent component labeling for artifact removal in fMRI

    Neuroimage

    (2008)
  • A. Weissenbacher et al.

    Correlations and anticorrelations in resting-state functional connectivity MRI: a quantitative comparison of preprocessing strategies

    Neuroimage

    (2009)
  • M.W. Woolrich et al.

    Bayesian analysis of neuroimaging data in FSL

    Neuroimage

    (2009)
  • J. Zhang et al.

    Evaluation and optimization of fMRI single-subject processing pipelines with NPAIRS and second-level CVA

    Magn. Reson. Imaging

    (2009)
  • Q.-H. Zou et al.

    An improved approach to detection of amplitude of low-frequency fluctuation (ALFF) for resting-state fMRI: fractional ALFF

    J. Neurosci. Methods

    (2008)
  • P. Zu Eulenburg et al.

    Meta-analytical definition and functional connectivity of the human vestibular cortex

    Neuroimage

    (2012)
  • H. Akaike

    A new look at the statistical model identification

    IEEE Trans. Autom. Control

    (1974)
  • B. Biswal et al.

    Functional connectivity in the motor cortex of resting human brain using echo-planar MRI

    Magn. Reson. Med.

    (1995)
  • B.B. Biswal et al.

    Toward discovery science of human brain function

    Proc. Natl. Acad. Sci. U. S. A.

    (2010)
  • E.T. Bullmore et al.

    Methods for diagnosis and treatment of stimulus-correlated motion in generic brain activation studies using fMRI

    Hum. Brain Mapp.

    (1999)
  • Cited by (1233)

    • Machine Learning of Functional Connectivity to Biotype Alcohol and Nicotine Use Disorders

      2024, Biological Psychiatry: Cognitive Neuroscience and Neuroimaging
    View all citing articles on Scopus
    View full text