Elsevier

NeuroImage

Volume 37, Issue 1, 1 August 2007, Pages 90-101
NeuroImage

A component based noise correction method (CompCor) for BOLD and perfusion based fMRI

https://doi.org/10.1016/j.neuroimage.2007.04.042Get rights and content

Abstract

A component based method (CompCor) for the reduction of noise in both blood oxygenation level-dependent (BOLD) and perfusion-based functional magnetic resonance imaging (fMRI) data is presented. In the proposed method, significant principal components are derived from noise regions-of-interest (ROI) in which the time series data are unlikely to be modulated by neural activity. These components are then included as nuisance parameters within general linear models for BOLD and perfusion-based fMRI time series data. Two approaches for the determination of the noise ROI are considered. The first method uses high-resolution anatomical data to define a region of interest composed primarily of white matter and cerebrospinal fluid, while the second method defines a region based upon the temporal standard deviation of the time series data. With the application of CompCor, the temporal standard deviation of resting-state perfusion and BOLD data in gray matter regions was significantly reduced as compared to either no correction or the application of a previously described retrospective image based correction scheme (RETROICOR). For both functional perfusion and BOLD data, the application of CompCor significantly increased the number of activated voxels as compared to no correction. In addition, for functional BOLD data, there were significantly more activated voxels detected with CompCor as compared to RETROICOR. In comparison to RETROICOR, CompCor has the advantage of not requiring external monitoring of physiological fluctuations.

Introduction

Over the last decade, blood oxygenation level-dependent (BOLD) and perfusion-based functional magnetic resonance imaging (fMRI) have become indispensable tools for studies of the working brain. When utilized together, the BOLD and perfusion signals can provide a quantitative understanding of the metabolic response to neural activity and provide insight into neurovascular coupling mechanisms (Hoge et al., 1999). However, as the fMRI community has moved to higher field strengths, physiological noise has become an increasingly important confound limiting the sensitivity and the application of fMRI studies (Kruger and Glover, 2001, Liu et al., 2006).

Physiological fluctuations have been shown to be a significant source of noise in BOLD fMRI experiments, with an even greater effect in perfusion-based fMRI utilizing arterial spin labeling (ASL) techniques (Kruger and Glover, 2001, Restom et al., 2006). Physiological sources of noise primarily include cardiac pulsations and respiratory-induced modulations of the main magnetic field. Additional sources include blood flow changes coupled to end-tidal C02 and vasomotion occurring at 0.1 Hz (Hu et al., 1995, Dagli et al., 1999, Glover et al., 2000a).

Approaches to removing cardiac and respiratory related noise include temporal filtering, image-based retrospective correction (RETROICOR), k-space-based correction (RETROKCOR) and navigator echo-based correction (DORK) (Hu et al., 1995, Biswal et al., 1996, Josephs et al., 1997, Glover et al., 2000a, Pfeuffer et al., 2002). More recently, RETROICOR has been extended to a general linear model (GLM) framework (Lund et al., 2006) and modified for use in ASL studies (Restom et al., 2006). A recent adaptation for BOLD-based imaging employs additional regressors describing variations in respiratory volume (Birn et al., 2006).

An alternate approach to the use of external measures of physiological activity or specially modified pulse sequences is to globally subtract average time courses from regions unlikely to be associated with neural activity, such as ventricles and large vessels (Petersen et al., 1998, Lund and Hanson, 2001). However, since this technique removes only the average time series, it is unable to account for voxel-specific phase differences in the noise due to physiological fluctuations. Additionally, component-based techniques, utilizing independent component analysis (ICA) or principal component analysis (PCA), have shown potential in identifying spatial and temporal patterns of structured noise (Thomas et al., 2002, McKeown et al., 2003, Beckmann and Smith, 2004). However, the utility of component-based methods has been limited to BOLD studies with sampling times short enough to clearly differentiate cardiac and respiratory elements from evoked responses (Thomas et al., 2002), in which case a temporal band pass filter would be adequate for noise removal.

In this paper we present and characterize a novel component-based method (CompCor) for the correction of physiological noise in BOLD and perfusion-based fMRI. We show that principal components derived from noise regions-of-interest (ROI) are able to accurately describe physiological noise processes in gray matter regions. In our presentation we investigate the use of two different methods for determining noise ROIs. The first method uses anatomical data to identify white matter and CSF voxels, while the second method uses the temporal standard deviation (tSTD) of the time series data to identify voxels dominated by physiological noise. We show that the use of principal components derived from a noise ROI as nuisance regressors in a GLM of the fMRI signal can significantly reduce the temporal standard deviation in resting-state scans and increase the sensitivity of functional BOLD and perfusion-based studies.

Section snippets

CompCor algorithm

The underlying assumption in the CompCor algorithm is that signal from a noise ROI can be used to accurately model physiological fluctuations in gray matter regions. The term “noise ROI” refers to areas (e.g., white matter, ventricles, large vessels) in which temporal fluctuations are unlikely to be modulated by neural activity and are primarily a reflection of physiological noise. The ability to model gray matter physiological noise elements is then predicated on the similarity between

Experimental protocol

Ten healthy subjects (ages 23 to 39) participated in the study after giving informed consent. Each subject viewed one periodic single trial visual stimulus consisting of a 20-s initial “off” period followed by 5 cycles of a 4-s “on” period and a 40-s “off” period. In addition to a periodic design, each subject viewed one block design consisting of 4 cycles of a 20-s “on” period and a 40-s “off” period. During the “on” periods, a full-field, full contrast radial 8-Hz flickering checkerboard was

Results

Fig. 5 shows normalized power spectra of physiological components estimated by RETROICOR, aCompCor, and tCompCor for the resting-state BOLD run of Subject 1. The top row (panel a) depicts the respiratory and cardiac components identified with the use of RETROICOR. Cardiac and respiratory peaks are located prominently at ∼ 1.2 Hz and ∼ 0.2 Hz, respectively. Panels b and c show the average spectra of elements estimated by aCompCor and tCompCor, respectively. Both variants of CompCor estimate

Discussion

In this paper, we have examined whether signal components derived from regions of interest that are unlikely to be modulated by neural activity can be used to estimate noise components (due to physiological fluctuations, subject motion, etc.) within activated regions. We considered two methods for the determination of the noise ROIs: (1) anatomical identification of significant areas of CSF and white matter and (2) definition of noise regions based upon their temporal standard deviation. We

Conclusion

We have shown that application of CompCor to ASL and BOLD fMRI time series can significantly reduce noise due to physiological fluctuations and other sources, such as subject motion. CompCor does not require external monitoring and can be applied in an automated fashion to reduce the confounding effect of physiological fluctuations on fMRI time series.

Acknowledgments

The authors thank Peter Costandi and Joanna Perthen for their valued assistance with the preparation of this paper. This work was supported in part by a Biomedical Engineering Research grant from the Whitaker Foundation and by NIH Grant R01NS051661.

References (29)

Cited by (2958)

View all citing articles on Scopus
View full text