1.

Introduction

Near-infrared spectroscopy (NIRS) is a developing and promising technology that monitors brain activity noninvasively and cheaply by measuring cerebral hemodynamic responses.^1,2 Because NIRS places fewer limitations on the subjects and environment than other functional neuroimaging methods such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET),^3,4 it has been widely applied in the cognitive and clinical research field.^5,6 However, the current application of NIRS in functional brain imaging is just the tip of the iceberg; the potential is far from fully exploited.^7,8 Current NIRS studies usually analyze only local changes in brain regions, with limited characterization of the brain’s functional architecture. The development of data analysis methods will be a critical factor for promoting the application of NIRS technology.^7,9

NIRS can provide high temporal resolution and reasonable spatial resolution in comparison to other traditional neuroimaging methods.¹⁰ With the development of multichannel NIRS systems, it has become possible to measure interactions between brain regions other than traditional local brain activation to derive NIRS-based connectivity.^8,11 NIRS-based connectivity is a novel analysis tool for NIRS data from the perspective of functional integration,¹² which could be complementary to NIRS activation analysis.

Executive control, which refers to the ability to attend to relevant information, ignore distracting information, overcome conflict, and select the appropriate response, is a key cognitive function of human beings. The Stroop task is a prominent technique for measuring executive control during conflict.¹³ Existing functional brain imaging studies demonstrate that the prefrontal cortex (PFC) is a predominant region involved in the Stroop task, which has functional hemispheric asymmetry.^14,15 However, current NIRS studies about hemispheric asymmetry in the Stroop task have focused on differences in activation, which can reveal only limited information about functional architecture of the brain.

In the present study, we aimed to fully examine the functional asymmetry of the PFC in executive control by employing traditional activation and NIRS-based connectivity analyses simultaneously. A color-word matching Stroop task that primarily activates the PFC was used.¹⁴ Hemodynamic signals from the PFC were recorded by NIRS. Wavelet transform coherence (WTC) analysis was employed to assess the intrahemispheric functional connectivity for each side of the PFC. Granger causality (GC) analysis was used to evaluate effective connectivity between hemispheres. The relationship between behavioral performance and brain activity was examined through Spearman correlation analysis.

2.

Method

2.2.

Procedures

A block design color-word matching Stroop task was employed in this study. Each stimulus consisted of two characters shown on a screen. The participants were asked to decide whether the meaning of the lower Chinese character was consistent with the color of the upper Chinese character [Fig. 1(a)]. If answer was “No,” the participants pressed a button using the right index finger, and if answer was “Yes,” they pressed the other button using the left index finger. Two types of stimulus conditions were displayed: incongruent and neutral. In the incongruent stimuli condition, the upper character was a color word (红, 黄, 蓝, 绿, 紫, meaning “red,” “yellow,” “blue,” “green,” and “purple”) printed in a disparate color, and the lower character was a color word presented in white. In the neutral stimuli condition, the upper character was a noncolor word (涂, 贯, 华, 球, 奖, meaning “scrawl,” “pass through,” “China,” “ball,” and “prize”) presented in red, yellow, blue, green, or purple [Fig. 1(a)]. For each stimulus condition, the trials were semirandomly mixed to avoid the consecutive appearance of more than three “Yes” or “No” trials.

Fig. 1

(a) The task sequence in a run (bottom) and examples of the two stimulus conditions (top). R, 60 s rest; B, 15 s baseline; N, neutral block; I, incongruent block and (b) the NIRS probe locations. The numbers (R1–R8 and L1–L8) at the midpoints of the source-detector pairs denote the NIRS channels. Channel L3 was over the F3 electrode position, and Channel R3 was over the F4 electrode position.

The experiment was composed of four runs, and each run was composed of four neutral and four incongruent task blocks [Fig. 1(a)]. There were 20 trials in each block, and each trial contained a stimulus shown for 1200 ms and a blank screen that lasted 300 ms, hence each block lasted 30 s. There was a 15-s rest period between blocks, a 60-s rest period between runs, and a 60-s rest period before the first run and after the last run. A sound was played to cue subjects 1 s before each block. There was a practice session for each subject before the formal experiment.

3.

Results

3.1.

Activation Results

Figure 2 displays the grand average hemoglobin signal for the two task conditions at one typical NIRS channel. Figure 3 shows the grand average activation responses for the left and right hemispheres. The activation responses were greater for the incongruent task compared with the neutral task for both hemispheres, but only the left hemisphere showed a significant Stroop effect for ${\mathrm{HbO}}_2$ signal. Before all the paired t-tests in this study, data were tested being normally distributed using Lilliefors test.

Fig. 2

The grand average hemoglobin signal for the two task conditions at NIRS channel L6. The gray area indicates standard error (SE). The gray vertical lines at 0 and 30 s indicate the start and end of the task. The hemoglobin changes were expressed in arbitrary units (A.U.).

Fig. 3

The grand average activation responses on (a) ${\mathrm{HbO}}_2$ signal and (b) Hb signal. Data are shown as $\text{mean}\text{values}\pm \mathrm{SE}$. N, neutral task; I, incongruent task. Black asterisk indicates $p<0.05$, FDR correction.

3.2.

Functional Connectivity Results

Figure 4(a) shows the intrahemispheric functional connectivity matrices on ${\mathrm{HbO}}_2$ signal averaged across subjects. Every matrix element was the coherence value for a NIRS channel pair (row coordinate and column coordinate). The mean value of the matrix elements was calculated for each matrix (except for elements of the secondary diagonal, which was the self-coherence of the channels) to indicate intrahemispheric functional connectivity for each hemisphere and each task [see Fig. 4(b)]. The intrahemispheric functional connectivities for both sides of the PFC were stronger for the incongruent task than for the neutral task, but only the left intrahemispheric functional connectivity showed a significant Stroop effect. In addition, for the incongruent or neutral tasks itself, the intrahemispheric functional connectivity of the left PFC was significantly stronger compared with that of the right PFC. Figure 5(a) shows the intrahemispheric functional connectivity matrices on Hb signal averaged across subjects, and Fig. 5(b) shows the grand average intrahemispheric functional connectivity on Hb signal. The left intrahemispheric functional connectivity on Hb signal showed a marginally significant Stroop effect. Fisher’s z-transform was employed for coherence values before the paired t-test.¹⁹

Fig. 4

(a) The average intrahemispheric functional connectivity matrices on ${\mathrm{HbO}}_2$ signal across subjects and (b) the grand average intrahemispheric functional connectivity on ${\mathrm{HbO}}_2$ signal. Data are shown as $\text{mean}\text{values}\pm \mathrm{SE}$. N, neutral task; I, incongruent task. Black asterisk indicates $p<0.05$, FDR correction.

Fig. 5

(a) The average intrahemispheric functional connectivity matrices on Hb signal across subjects and (b) the grand average intrahemispheric functional connectivity on Hb signal. Data are shown as $\text{mean}\text{values}\pm \mathrm{SE}$. N, neutral task; I, incongruent task. Gray asterisk indicates $p<0.1$, FDR correction.

3.3.

Correlation between Response Time and Activation/Functional Connectivity

The results of correlation showed that there was a greater negative correlation between RT and the left PFC, compared with the right PFC (see Fig. 6). The negative correlation indicated that the stronger the brain activation/functional connectivity, the better the behavioral performance. Besides, both the correlations between RT and left intrahemispheric functional connectivity on ${\mathrm{HbO}}_2$ signal and Hb signal were higher than activation.

Fig. 6

The correlation between RT and activation for both sides of the PFC on (a) ${\mathrm{HbO}}_2$ signal and (c) Hb signal. The correlation between RT and intrahemispheric functional connectivity for both sides of the PFC on (b) ${\mathrm{HbO}}_2$ signal and (d) Hb signal.

3.4.

Effective Connectivity Results

Figure 7 shows that the average ΔDOI values across subjects were positive for all time lags (subject 9 was excluded from the analysis for the 1-s time lag on ${\mathrm{HbO}}_2$ signal because the values were more than three standard deviations from the mean value), and there were marginally significant differences between the incongruent and the neutral tasks for the 1-s time lag on ${\mathrm{HbO}}_2$ signal and for the 1.5-s time lag on Hb signal. The effective connectivity results indicated that the flow of information from the left to the right PFCs increased accompanying conflict processing.

Fig. 7

The effective connectivity results. (a) The average ΔDOI values on ${\mathrm{HbO}}_2$ signal and (b) the average ΔDOI values on Hb signal. Gray asterisk indicates $p<0.1$, FDR correction.

4.

Discussion

This study employed traditional activation and novel NIRS-based connectivity analyses simultaneously to examine the functional asymmetry of the PFC in executive control. The results of both the NIRS activation and the NIRS-based connectivity analyses demonstrated a leftward lateralization for the Stroop effect and the correlation with behavioral performance. However, NIRS-based connectivity was more sensitive for hemispheric lateralization identification than activation analysis. Furthermore, NIRS-based connectivity can reveal and model the interaction between hemispheres, helping us to understand the functional organization of hemispheric asymmetry in the Stroop task more comprehensively. The results on ${\mathrm{HbO}}_2$ and Hb are in consistency, with slightly better for results on ${\mathrm{HbO}}_2$. Previous studies have demonstrated that, compared with Hb, ${\mathrm{HbO}}_2$ had a better signal-to-noise ratio (SNR) and was the more sensitive indicator of changes in regional cerebral blood flow.^24,25 And the correlation between BOLD and ${\mathrm{HbO}}_2$ was higher than Hb, which may because of the higher SNR of ${\mathrm{HbO}}_2$.²⁶

Compared with the neutral task, the activation responses were greater for the incongruent task for both sides of the PFC, and the left hemisphere showed a significant Stroop effect. The activation results indicated that the left PFC is dominant for Stroop interference processing, which is consistent with previous fMRI studies,¹⁴ indicating that the design of this study was correct.

The functional connectivity results also demonstrated a left hemisphere dominance for Stroop interference processing. In addition to the Stroop effect, for the incongruent or the neutral tasks, there was also a leftward lateralization, which is in line with the verbal nature of the two types of tasks.²⁷ Previous studies have shown that, for right-handed subjects, the left hemisphere is more functionally specialized for verbal tasks than the right hemisphere.²⁸

The results of both types of correlation analysis (RT and brain activation and RT and functional connectivity) indicated that the left PFC plays a primary role in affecting behavioral performance, in accordance with a previous study that reported poorer performance on the Stroop task in patients with left PFC lesions.²⁹ Compared with activation analysis, intrahemispheric functional connectivity of the left PFC had a greater correlation with RT, indicating that the functional connectivity could be more sensitive for identifying the primary brain regions that affect behavioral performance. The results also suggested that the functional connectivity could predict behavioral performance better than activation.

The reason that functional connectivity was more sensitive than activation for identifying hemispheric lateralization and predicting behavioral performance may be that there are uncertainties in quantifying hemoglobin change using the MBLL method.⁷ Although some previous researches reported results in μM change,⁷ it is not quantitatively accurate because the partial volume effect was not considered.^30,31 Additionally, DPF varies between different investigated areas and subjects. These two factors lead to errors in measured activation. However, the trend of hemoglobin change, on which functional connectivity analysis is based, is still correct.⁷

Executive control over conflict includes two steps: conflict detection and resolution.¹³ The results of the effective connectivity analysis demonstrated increased information flow from the left to the right PFCs for the incongruent task versus the neutral task, indicating a leading role of the left PFC in regard to the right PFC during conflict processing. The results seem to show that the left PFC detected the conflict first and then aroused the right PFC for conflict resolution. The results support the viewpoint that the left PFC is involved in conflict detection, which is in line with a previous fMRI study that suggested that conflict detection is achieved by the left PFC.³² The effective connectivity results illustrate that the NIRS-based brain connectivity can deliver more comprehensive information compared with traditional activation.

A meta-analysis about lateralized Stroop studies demonstrated that, although some studies showed a leftward lateralization, there are no significant differences between the hemispheres on the whole.³³ But the studies included in this meta-analysis were unilateral presentation Stroop experiments, and just employed reaction time as the dependent measure. Both the experimental paradigm and assessment indicator are different from ours, and these differences may lead to different conclusions.³³ What is more, instead of brain imaging technology, these studies used behavioral performance to assess the brain activity indirectly; thus, this assessment may be inappropriate. In the future, various task paradigms should be considered to further investigate the hemispheric lateralization during Stroop tasks. And multiple parameters, including brain activity and behavioral performance, should be obtained.

In addition, right dominance of PFC activity can be caused by the mental stress responses of the autonomic nervous system during mental stress tasks.³⁴ In this study, there was no mental-related information in the experiment materials. Before the formal experiment, there were practice sessions to familiarize the subjects with the task. Subjects reported that they were relaxed and peaceful during the formal experiment. Left dominance of PFC activity was observed in this study, indicating that the hemispheric lateralization was indeed caused by cognitive function.

In this study, due to the absence of reference channels, some approaches for reducing the interference of superficial layers, such as adaptive filtering technique, cannot be used here. For blind source separation, an estimation of the eigenstructure based on the prior information about the degree of noise or visual inspection with the step-by-step removal of eigenvectors is required. Then, subjective errors are likely introduced in the case that no reference channel is included.³⁵ In addition, the extent of the functional region is comparable with the spatial coverage of the probe in our study; PCA filters can negatively affect the estimate of the hemodynamic response.³⁶ It will be a similar situation for using global regression to reduce the influence. But what we focus on is the differences of task conditions in an experiment. And the evaluation of the differences itself would counteract the influence of task-unrelated signal components.³⁷

5.

Conclusion

This study analyzed hemispheric asymmetry using NIRS from multiple perspectives, including activation, functional connectivity, effective connectivity, and behavioral analysis. The results indicate that, compared with traditional activation, NIRS-based connectivity is more sensitive and delivers more comprehensive information that can help us to better understand the functional organization of the brain. This study demonstrates that, as a novel NIRS analysis tool, NIRS-based connectivity can exploit the capabilities of NIRS effectively, helping to promote the application of NIRS in neuroscience research, especially in clinical study.