Combining independent component analysis and correlation analysis to probe interregional connectivity in fMRI task activation datasets

https://doi.org/10.1016/S0730-725X(00)00190-9Get rights and content

Abstract

A new approach in studying interregional functional connectivity using functional magnetic resonance imaging (fMRI) is presented. Functional connectivity may be detected by means of cross correlating time course data from functionally related brain regions. These data exhibit high temporal coherence of low frequency fluctuations due to synchronized blood flow changes. In the past, this fMRI technique for studying functional connectivity has been applied to subjects that performed no prescribed task (“resting” state). This paper presents the results of applying the same method to task-related activation datasets. Functional connectivity analysis is first performed in areas not involved with the task. Then a method is devised to remove the effects of activation from the data using independent component analysis (ICA) and functional connectivity analysis is repeated. Functional connectivity, which is demonstrated in the “resting brain,” is not affected by tasks which activate unrelated brain regions. In addition, ICA effectively removes activation from the data and may allow us to study functional connectivity even in the activated regions.

Introduction

Functional magnetic resonance imaging (fMRI) time-series data from functionally connected regions of the brain exhibit high temporal coherence of low frequency fluctuations (≤0.1 Hz) [1] in subjects performing no prescribed task (“resting state”). This effect is a result of a combination of interrelated electrochemical and physiological brain processes. During a “resting” state, there is spontaneous firing of cortical neurons. This spontaneous neuronal activation is always followed by changes in regional cerebral blood flow [2]. Such blood flow fluctuations change blood oxygen levels and therefore the blood oxygen level dependent (BOLD) signal [3], [4], [5]. Due to interregional neural connectivity, activation in a specific brain region affects remotely located neurons in other areas through the efferent output. By cross correlating the signal time course from an ROI in the brain with the time courses generated in all other voxels, it may be possible to detect areas functionally connected to the initially selected ROI1. This concept has been studied extensively in subjects instructed to refrain from any cognitive, sensory or motor activity [1], [6], [7]. However, even if we assume that absolutely no motor activity is taking place during a “resting” condition it is almost impossible to cancel all cognitive or sensory events. Therefore, studying functional connectivity in “resting” state studies is not ideal since uncontrollable activations can alter the final cross correlation results. On the other hand, detecting functional connectivity may be more advantageous than the conventional task-induced activation analysis in that there is no dependence of the results on type of task, subject’s performance, or degree of involvement of various brain regions to specific tasks [7].

In this paper we study functional connectivity by applying the cross correlation analysis to task-related activation datasets. This is an attempt to measure functional connectivity more effectively by having more control over the subject’s cognitive activities compared to “resting” acquisitions. However, using this approach the brain regions that can be studied for functional connectivity are limited to only those that are not affected by the performance of the task. Therefore, we implemented and tested a method to remove the effect of task activation from the data using independent component analysis [8] (ICA). Cross correlation analysis results were then compared in “resting” state and task activation studies before and after removal of the activation.

Section snippets

Methods

This research was performed on a clinical 1.5T GE Horizon MRI scanner (General Electric, Waukesha, WI) with high-speed gradients. Six right-handed volunteers, between 20–40 years of age, possessing no known neurologic disorders participated in this study. All subjects signed a consent in accordance with institutional policy. A prototype RF quadrature birdcage head coil (Medical Advances Inc., Milwaukee, WI) was used for signal transmission and reception. Aircraft-type earphones with additional

Results

Z-score maps showed that bilateral finger tapping activated the left and right primary motor cortices, primary sensory cortices, cerebellum and supplementary motor areas (SMA). Activation in the thalamus and the putamen was inconsistent across subjects. As expected, z-score maps demonstrated that passive listening to narrated text resulted in bilateral activation of the primary auditory cortex. A flashing checkerboard produced bilateral activation in primary visual cortex and in motion

Discussion

Along with other anatomic, histochemical, and physiological techniques [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], there are indications that cross correlation analysis on “resting” state MRI data may be a valid method for studying functional connectivity in the human brain. This method may be more advantageous than typical task-induced activation analysis since the latter may detect only a subset of a specific neural system or underestimate the size and number of

Conclusion

The present study demonstrated that interregional functional connectivity of the human brain could be studied in task-induced activation datasets as opposed to the conventional “resting” state acquisitions. Cross correlation analysis of the activation data includes all the advantages of the same analysis applied on “resting” state data and additionally may avoid the effects of unidentified, uncontrolled neural activations. Moreover, using task activation data to study functional connectivity

References (32)

  • M.J Lowe et al.

    Functional connectivity in single and multislice echoplanar imaging using resting-state fluctuations

    Neuroimage

    (1998)
  • G.L Gerstein et al.

    Identification of functionally related neural assemblies

    Brain Res

    (1978)
  • B Biswal et al.

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

    Magn Res Med

    (1995)
  • E.V Golanov et al.

    Spontaneous waves of cerebral blood flow associated with a pattern of electrocortical activity

    Am J Physiol

    (1994)
  • S Ogawa et al.

    Intrinsic signal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonance imaging

    Proc Natl Acad Sci USA

    (1992)
  • K.K Kwong et al.

    Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation

    Proc Natl Acad Sci USA

    (1992)
  • Jezzard P, LeBihan D, Cuenod D, Pannier L, Prinster A, Turner R. An investigation of the contribution of physiologic...
  • J Xiong et al.

    Interregional functional connectivity to primary motor cortex revealed using MRI resting state images

    Hum Brain Mapp

    (1999)
  • P Comon

    Independent component analysisa new concept?

    Signal Processing

    (1996)
  • M.J Lowe et al.

    Spatially filtering functional magnetic resonance imaging data

    Magn Res Med

    (1997)
  • P.A Bandettini et al.

    Processing strategies for time-course data sets in functional MRI of the human brain

    Magn Res Med

    (1993)
  • M.J Lowe et al.

    Treatment of baseline drifts in fMRI time series analysis

    J Comput Assist Tomogr

    (1999)
  • M.J McKeown et al.

    Analysis of fMRI data by blind separation into independent spatial components

    Hum Brain Mapp

    (1998)
  • M.J McKeown et al.

    Spatially independent activity patterns in functional MRI data during the stroop color-naming task

    Proc Natl Acad Sci USA

    (1998)
  • B Biswal et al.

    Blind source separation of multiple signal sources of fMRI data sets using independent component analysis

    J Comp Assist Tomogr

    (1999)
  • M.J McKeown et al.

    Independent component analysis of fMRI dataexamining the assumptions

    Hum Brain Mapp

    (1998)
  • Cited by (167)

    View all citing articles on Scopus
    View full text