INTPSY-10996; No of Pages 9 International Journal of Psychophysiology xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho

Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study Ling Wei a, Yingjie Li a,⁎, Xiaoli Yang b, Qing Xue c, Yuping Wang c a b c

School of Communication and Information Engineering, Shanghai University, Shanghai, PR China Department of Electrical & Computer Engineering, Purdue University Calumet, Hammond, IN, United States Department of Neurology, Xuanwu Hospital, Capital Medical University, Beijing, PR China

a r t i c l e

i n f o

Article history: Received 1 December 2014 Received in revised form 12 April 2015 Accepted 28 May 2015 Available online xxxx Keywords: Mild cognitive impairment EEG Small-world network Complex network analysis Phase synchronization Time-evolution

a b s t r a c t The present study evaluated the topological properties of whole brain networks using graph theoretical concepts and investigated the time-evolution characteristic of brain network in mild cognitive impairment patients during a selective attention task. Electroencephalography (EEG) activities were recorded in 10 MCI patients and 17 healthy subjects when they performed a color match task. We calculated the phase synchrony index between each possible pairs of EEG channels in alpha and beta frequency bands and analyzed the local interconnectedness, overall connectedness and small-world characteristic of brain network in different degree for two groups. Relative to healthy normal controls, the properties of cortical networks in MCI patients tend to be a shift of randomization. Lower σ of MCI had suggested that patients had a further loss of small-world attribute both during active and resting states. Our results provide evidence for the functional disconnection of brain regions in MCI. Furthermore, we found the properties of cortical networks could reflect the processing of conflict information in the selective attention task. The human brain tends to be a more regular and efficient neural architecture in the late stage of information processing. In addition, the processing of conflict information needs stronger information integration and transfer between cortical areas. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Mild cognitive impairment (MCI), characterized by cognitive decline more severe than expected in normal aging, is considered as the preclinical state between normal elderly cognition and dementia (Gauthier et al., 2006; Portet et al., 2006; Winblad et al., 2004). Individuals with MCI often have mild problems performing complex functional activities, and show a fourfold increased risk for the development of Alzheimer's disease (AD) when compared to healthy elderly people (Albert et al., 2011; Petersen, 2004). In recent years, studies in the general population suggest that AD or MCI may be associated with a deficit in selective attention early in the progression of the disease (Baddeley et al., 2001; Belleville et al., 2007; Fernandez-Duque and Black, 2006; Ko et al., 2005; Krinsky-McHale et al., 2008; Pignatti et al., 2005). Selective attention includes the ability to ignore perceptual distracters and the ability to withhold responses to goals. It reflects the top-down control (from frontal and parietal areas) of information processing based on task demands (Booth et al., 2005; Bressler et al., 2008; Geerligs et al., 2014). Furthermore, this modulation processing relies on long-range inputs from and interactions with a network of multiple regions (Gazzaley ⁎ Corresponding author at: P.O. Box 98, 99, Shangda Road, Baoshan District, Shanghai University, Shanghai 200444, PR China. Tel./fax: +86 21 66137258. E-mail address: [email protected] (Y. Li).

and Nobre, 2012; Zhang et al., 2014). In addition, the functional disconnection of brain regions in MCI and AD has been proved by many studies (Bai et al., 2008; Greicius et al., 2004; Jiang et al., 2014; Liu et al., 2012; Rombouts et al., 2005). In recent years, the complex network analysis based on graph theoretical approaches has brought a method with a number of neurobiological meaningful and easily computable measures to the investigation of brain functional networks (Bullmore and Sporns, 2009; Deuker et al., 2009; He et al., 2007; Stam et al., 2007a, 2007b). An ordered network has a high clustering coefficient C (a measure that depicts the connectedness of immediate neighbors around individual vertices) and a long characteristic path length L (an index reflecting the overall integration of the network). In contrast, randomly organized networks are characterized by a low C and a short L. The small-world network, characterized by a high degree of clustering and short path length linking individual network nodes, has been an attractive model for the description of complex brain networks (Achard and Bullmore, 2007; Chen et al., 2008; Wu et al., 2012). Researchers found that both the anatomical and functional brain networks are small-world network (Bassett and Bullmore, 2006; Stam et al., 2007a, 2007b). Furthermore, comparisons of topologies between subject populations appear to reveal connectivity abnormalities in neurological and psychiatric disorders. The AD and MCI patients, for example, show abnormal properties of cortical networks and loss of small-world characteristics in recent functional magnetic resonance imaging findings,

http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015 0167-8760/© 2015 Elsevier B.V. All rights reserved.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

2

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

which reported that the local and global functional connectivity disruptions were found in these patients (Liu et al., 2012; Yao et al., 2010). In specific, AD patients showed the longest L and the largest C, and MCI exhibited intermediate values compared with normal controls. However, the whole-brain functional networks especially their transient activities during selective attention task of MCI are not yet fully understood. Most studies show that the cortical neural synchronization of oscillations has been used to evaluate the brain functional network (Stam and van Straaten, 2012). And the scalp electroencephalogram (EEG) representing directly the ongoing neural activity, is appropriate for the investigation of networks related to various oscillatory frequency bands. The related studies on brain oscillations have revealed that the alpha and beta oscillation was correlated with cognitive decline (Bian et al., 2014; Jeong, 2004; Uhlhaas and Singer, 2006). The change of neural synchronous oscillation can be estimated with linear measures such as coherence and nonlinear techniques based on phase relationships. Relative to coherence, phase synchronization which overcome the confounding effects of mixing amplitude and phase and refers to the interdependence between the instantaneous phases of two EEG signals and, has been interpreted as a susceptive method to measure functional synchronization of EEG data (Basar et al., 2010; Bressler and Menon, 2010; Sun and Small, 2009). Related studies demonstrated a general decrease of phase synchrony in correlation with cognitive decline and AD (Knyazeva et al., 2010; Koenig et al., 2005). Compared with age matched control subjects, MCI and AD patients showed decreased global phase EEG synchrony values in alpha, beta, and gamma frequency bands, and increased global phase EEG synchrony values in the delta band in resting condition (Koenig et al., 2005; Park et al., 2008). In graph theory analysis, the brain function connectivity network can be constructed using the interaction between signals of all pair-wise combinations of the channels, such as correlation coefficients (Ahmadlou et al., 2014) and synchronization likelihood (Ponten et al., 2007). The phase synchronization analysis, which can separate the phase component from the amplitude component, is considered to be a valid method to construct and evaluate the graph parameters in specific oscillatory frequency bands (Wang et al., 2012; Wu et al., 2012). In this study, we used phase synchronization index to measure the cortical neural synchronization of oscillations in alpha and beta bands and construct the brain functional networks for MCI patients and healthy controls. Then, the graph theoretical analysis was applied to study the network characteristics in different information process. Our hypotheses were as follows: (1) the small-world characteristic in patients with MCI was lost. (2) In a selective attention task, MCI patients had different network parameters compared with normal controls. 2. Methods 2.1. Participants All the subjects were recruited from community by Xuanwu Hospital of Capital Medical University according to MMSE (mini-mental state examination) and MoCA (Montreal cognitive assessment) criteria. Ten human subjects diagnosed with mild cognitive impairment (in the following we use the term “MCI patients”) and seventeen elderly healthy control subjects with no history of psychiatric disorder participated in this study. Table 1 showed the socio-

demographic characteristics of participants. All subjects had no personal neurological history, no drug or alcohol abuse, no current medication, and had normal or corrected-to-normal vision. Every subject signed an informed consent according to the guidelines of the Human Research Ethics Committee and was paid for participation after the experiment.

2.2. Stimuli and procedure Selective attention can be studied in visual and spatial tasks. In visual task, one method is to ask participants to ignore some stimulus dimensions (e.g., identity, location) and respond based on a target dimension (e.g., color). In this study, we asked the subjects to attend to color and ignore the shape of stimuli. The stimuli were some familiar colored shapes. Each stimulus had one of four colors (red, yellow, green, and white), and had one of four shapes (triangle, quadrangle, hexagon and round). The two stimuli in a pair were sequentially presented. The first stimulus and the second stimulus were presented on a monitor screen for 300 ms each with an inter-stimulus interval of 500 ms. Another trial began after 5 s ITI (inter-trial interval) (see Fig. 1). The stimuli were presented at the screen center of a computer-controlled monitor to each participant with a black background. The stimulus pairs were randomly presented 50 times in sequence and had equal probability. Subjects were seated at 1 m distance from the screen. They were instructed to sit quietly and to look at the cross in the center of the screen. The averaged visual angle of S1 and S2 was adjusted to 2.1°. In this task, participants were asked to discriminate whether the color of S2 was identical to that of S1 and ignored the shape of the stimulus. When the color of S2 was the same as that of S1, they were asked to press the left button of a push pad; when the color of S2 differed from S1, they pressed the right button. In statistic analysis, we divided the stimuli into two types (color match and color mismatch).

2.3. EEG recording and pre-processing The electroencephalogram (EEG) was recorded using a sixtyfour-channel EEG system with 60 surface electrodes mounted in an electrode cap (electrode impedance b5 kΩ, 0.05–100 Hz band pass, 1000 samples/s). Vertical and horizontal EOGs were simultaneously recorded to monitor eyes movement and blinks. The data were referenced to one electrode placed on nose. Trials with ocular, saccades artifacts and artifacts N± 100 μV were rejected before exporting. Artifact-free data were then segmented ranging from 200 ms before S1 to 1000 ms after S2 stimulus onset for all conditions. Moreover, we chose 5 time windows (200–0 ms before S1, 0– 200 ms, 200–400 ms, 400–600 ms, 600–800 ms after S2) for subsequent analysis in this study in order to explore the time-evolution of brain function network. In the following, each small epoch was decomposed into scale levels by wavelet packet decomposition (more details see reference (Akay, 1995)) in accordance with the traditional frequency bands in EEG analysis, i.e., beta: 16–32 Hz, alpha: 8–16 Hz. In the present work, the ‘db5’ mother wavelet which was better to reconstruct the original EEG signals from our prior study was employed with 7 levels decomposition.

Table 1 Mean values ± standard deviation of socio-demographic characteristics. Normal

MCI patients

Significance

10 4/6 75.20 ± 7.74 7.0 ± 3.9

\ \ 0.884 0.707

Controls Number of subjects female/male Age (years) Education (years)

17 7/10 74.82 ± 5.54 7.59 ± 3.86

2.4. Network analysis 2.4.1. Phase synchronization It is well known at present that the phases of two coupled nonlinear oscillators may synchronize even if their amplitudes remain uncorrelated. So we use phase synchronization to analyze the characteristics of EEG during a cognitive task.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

3

Fig. 1. A task flow diagram of the sample stimuli and the experimental design.

For a continuous-time signal, an analysis signal can be defined as follows (Rosenblum et al., 1996): H

Z x ðt Þ ¼ xðt Þ þ i^xðt Þ ¼ AHx ðt Þeiθx ðt Þ

ð1Þ

where the functions is the Hilbert transform of x(t): ^xðt Þ ¼

1 π

Z

þ∞

xðτÞ dτ: t−τ −∞

ð2Þ

Similarly, H

Z y ðt Þ ¼ AHy ðt Þeiθy ðt Þ :

ð3Þ

The phase synchrony index (PSI) γ for two instantaneous phases θx and θy is defined as D E   γ ¼  eiðnθx −mθy Þ ∈½0; 1

ð4Þ

where n and m are integers (usually n = 1 = m), and 〈·〉 denote average over time (Quian et al., 2002). In this study, the phase synchrony index was calculated between each possible pairs of EEG channels in alpha and beta frequency bands for five time windows. For each segment, a square 60 × 60 matrix could be obtained (60 was the number of EEG channels), and the matrix was defined with each element as the mean value of all segments. 2.4.2. Complex networks analysis According to graph theory, the brain can be seen as a complex functional network, which can be basically represented as graphs. The first step in applying graph theoretical analysis to PSI matrices is to convert the matrix into a binary undirected graph with a fixed degree K. Binary means edges between nodes either exist or do not exist. Once the matrix has been converted to a graph, the next step is to characterize the topological properties of the network between groups. The minimum path length between two nodes in a network, Lij, is the smallest number of edges that must be traversed to make a connection between node i and node j. The characteristic path length L of graph G, which is defined as the average of the shortest path lengths between two generic nodes, is thus a measure of the global connectivity of the network: L¼

X 1 L : i≠ j∈G i j NðN−1Þ

ð5Þ

where Gi is the subgraph of nodes and edges connected to node i, NGi is the number of edges in subgraph Gi, and Ljk is the minimum path length between nodes j and k in the subgraph. The clustering coefficient C, which is the average value of Ci all over the network, is a measure of the local connectivity of a regional node (Watts and Strogatz, 1998). In this study, we analyzed C and L as a function of degree K, which is the average number of edges per vertex. In this way, graphs in both groups are guaranteed to have the same number of edges so that any remaining differences in C and L between the groups reflect differences in graph organization. The values of C and L as a function of degree K were compared with the theoretical values of C and L for random (C = K / N, L = ln(N) / ln(K)) graphs. For each K, 20 random networks were generated, and the mean C and L were calculated. To establish the significance of network topological characteristic, we calculated the ratios C/Cs (γ) and L/Ls (λ) for two groups where Cs and Ls denote the values of C and L for matched random reference graphs. Small-world network organization is evident when values of C/Cs are significantly greater than 1 while the values of L/Ls are near the value of 1 (Micheloyannis et al., 2006; Stam and Reijneveld, 2007). This can be summarized by the scalar σ = γ/λ, which will be greater than unity for a small-world network (Humphries et al., 2006). The software instrument used here were Matlab for off-line calculation of PSI value and Brain Connectivity Toolbox (http://www.brainconnectivity-toolbox.net/) for the calculation of graph theoretical measures. 2.4.3. Statistical analysis Repeated measures analysis of variance (ANOVA) was performed to test the significant changes of network characteristics (C, L and σ). Two within-group factors: Stimulus (match and mismatch), Time (win1, win2, win3, win4, win5) and a between-group factor Group (the MCI patients and the normal control) were considered as fixed factors for each K (in this paper, 5 ≤ K ≤ 30). The simple effect analysis was performed if any interaction between factors was found. All analysis was conducted at the 0.05 level of significance. 3. Results A 2 (Group: MCI patients, normal controls) × 2 (Stimulus: match and mismatch) × 5 (Time: −200–0 ms, 0–200 ms, 200–400 ms, 400–600 ms, 600–800 ms) ANOVA revealed the main effect of Group, Time and interaction of Stimuli × Group, Time × Stimuli, Time × Stimuli × Group in alpha and beta bands (see Table 2). 3.1. Group effect of the cortical network

The local clustering coefficient of node i, Ci, is defined as follows: Ci ¼



1

 N NGi −1

X j;k∈N Gi

1 L jk

ð6Þ

Fig. 2 showed the C, L and σ as a function of degree K (from 5 to 30) in MCI patients and normal controls at resting state. Similar results were found in other times windows. Over the range of K investigated, the C

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

4

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

Table 2 The main effect and interaction effect of factors in network parameters (C, L and σ) in alpha and beta bands (P b 0.05). Frequency band

Network parameters

K range Group effect

Stimuli effect

Time effect

Interaction effect

Alpha

C

12–13,15–19

24,25

12–18

Beta

L σ C L σ

5–13 18–25 10–25 6–7,10–25 16–18,23–25

12–20 13–15 \ \ \

9,12 19,21 20–25 11–17 21–25

16–25 (Stimuli × Group) 10 (Time × Stimuli × Group) 24,25 (Stimuli × Time) 24 (Time × Stimuli × Group) \ 12 (Time × Stimuli × Group) 16 (Stimuli × Time)

remained relatively stable and the L decreased with a gradual increase of K. When the K increased, the σ in all the two networks fell down. In alpha and beta bands, both C and L in the MCI networks were lower than in normal controls. In this study, the two groups demonstrated small-world property (σ N 1) over an entire range of degree. Our finding that the σ decreased in MCI patients compared with normal controls suggested that MCI patients had a further loss of small-world attributes. In different time windows, the ANOVA of network parameters also revealed a significant main effect of Group in alpha and beta bands. More details can be seen in Table 3. In alpha band, the C, L and σ in MCI networks were significantly lower than in normal controls in −200–0 ms and 400–600 ms. In 0–600, we also found significant differences for C and L between MCI and controls. In 200–600 ms, the simple effect analysis of C showed the difference of group in alpha band was

only significant in mismatching task (win3: K = 16–30, win4: K = 21– 24). In 600–800 ms, significant Group effect was only found for the L of networks. In beta band, group difference in the two measures were significant in −200–800 ms. The network parameters (C and L) of MCI patients were close to that of random network. And the σ of MCI network was closer to unity both in alpha and beta bands.

3.2. Time effect of the cortical network In both alpha and beta bands, we found significant main effect of Time, interaction effect of Time × Stimuli and Time × Stimuli × Group for parameters of brain network (Table 3). The C, L and σ of brain network in − 200–0 ms were the largest than in other time windows.

Fig. 2. Mean clustering coefficients C (top row), characteristic path lengths L (second row) and σ (bottom row) as a function of degree K in time window 1 (−200–0 ms) between two groups (Normal controls and MCI patients) in alpha (left column) and beta (right column) band. K ranges from 5 to 30 with increments of 1. The C, L and σ in the MCI networks were lower than in normal controls in alpha and beta bands.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

5

Table 3 The K range of significant Group effect in five time windows (P b 0.05). Frequency band

Network parameters

Alpha

C L σ C L σ

Beta

K range −200–0 ms

0–200 ms

200–400 ms

400–600 ms

600–800 ms

12–19 6–16 18–24 11–18 11–12 \

8–11 5,7 \ 10–30 11–13,24–30 12–15,22–30

12–20 7–8 \ 11–30 9–25 13–30

12–17,19–20 6–12 17–20,22–25 10–14,17 5–6,13–14 16,18

\ 6,11–14 \ 6,7,11–14 22–25 \

Fig. 3 showed the Time effect in alpha and beta bands. More specifics of results in each time window were followed. In 0–200 ms, a significant main effect of Stimuli was found for K in the range of 10–14, 17–19 in alpha band and 12–14 in beta band. The L of network in matching task was higher than that in the mismatching task. We also found the interaction effect of Stimuli × Group for L in alpha band (K = 20– 30). Simple effect analysis found different effect of Stimuli for the two groups. In matching task, normal controls had shorter L while MCI patients had longer L than that in mismatching task. In 200– 400 ms, significant main effect of Stimuli (C: K = 17, 27–30, L: K = 24– 28, σ: K = 8–16) and interaction effect of Stimuli × Group (C: K = 16– 30) were found in alpha band. The network for two groups had longer L and smaller σ in matching task (Fig. 4A and E). And for MCI patients, the C of network in matching task was higher than that in the mismatching task. In 400–600 ms, we only found significant main effect of Stimuli for K in the range of 25–27 (Fig. 4B). In 600–800 ms, the

statistical analysis revealed main effect of Stimuli (C: K = 24–26) and interaction effect of Stimuli × Group (C: K = 24–30) in alpha band. The C was significant larger in matching task relative to the mismatching task only for MCI patients (K: 24–30). Furthermore, we also found the main effect of Stimuli (σ: K = 15–21, 26) and interaction effect of Stimuli × Group (L: K = 20–30) in beta band. The σ in mismatching task was larger than in matching task for MCI patients. Simple effect analysis suggested that the L of network in matching task was larger than in mismatching task for normal controls (see Fig. 4D). 4. Discussion In the present study, we explored the group difference of MCI patients and normal controls on EEG alpha and beta bands brain networks by measuring phase synchronization with graph theoretical tools in different cognitive process. Analysis revealed that the characteristic of

Fig. 3. Mean clustering coefficients C (top row), characteristic path lengths L (second row) and σ (bottom row) in five time windows (−200–0 ms, 0–200 ms, 200–400 ms, 400–600 ms, 600–800 ms) between two groups (Normal controls and MCI patients) in alpha and beta bands. The C, L and σ of brain network in −200–0 ms were the largest than in other time windows.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

6

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

Fig. 4. The difference of network characteristics between match stimuli and mismatch stimuli in alpha and beta bands. (A) The stimuli difference of L in alpha band in 200–400 ms. (B) The stimuli difference of L in alpha band in 400–600 ms. The L of MCI network was lower in matching task relative to the mismatching task. (C) The stimuli effect of C in beta band in 600– 800 ms. The C of network in matching task was smaller than in mismatching task. (D) The stimuli effect of L in beta band in 600–800 ms. The L in mismatching task was shorter than in matching task for normal controls. (E) The stimuli effect of σ in alpha band in 200–400 ms. The σ of network in matching task was larger than that in mismatching task. (F) The stimuli effect of σ in beta band in 600–800 ms. The σ of network in mismatching task was larger than that in matching task.

network in MCI patients was more close to the theoretical values of random network and small-world network characteristics was lost in MCI patients compared with healthy controls. Furthermore, we found similar time-evolution characteristic of cortical networks during the color selective attention task for two groups. 4.1. Abnormal properties of cortical networks in MCI patients during a selective attention task Previous studies have reported that the human brain has evolved into a complex but efficient neural architecture to maximize the power of information processing for the large clustering coefficients but short characteristic path lengths in whole brain networks. The L and C indicated global and local functional integration of whole brain networks that constitute basic cognitive processes (Sporns and Zwi, 2004). Previous studies that used resting-state functional magnetic resonance imaging (fMRI) found the brain networks of AD patients had longest characteristic path lengths and the largest clustering coefficients, while the properties of the MCI network exhibited intermediate values (Liu et al., 2012; Yao et al., 2010). But no significant differences in the two measures between normal controls and MCI patients were found in their studies. Meanwhile, a functional brain network study using fluorodeoxyglucose positron emission tomography (FDG-PET) found the clustering of networks was lower in MCI patients compared with normal controls, while path length was not different between the two groups (Seo et al., 2013). In this

paper, we detected that MCI patients showed smaller C and shorter L over a wide range of degree K than normal controls in both alpha and beta bands at resting state. These results reflect the decline in functional integration in whole-brain network (Liu et al., 2012; Yao et al., 2010) and disrupted information processing among distant brain regions across the whole brain (Liu et al., 2014). All these abnormal properties may be explained by the patterns of regional brain hypometabolism in MCI (Seo et al., 2013). The C and L as a function of degree K also demonstrated that compared to the network of normal controls, the characteristic of network in MCI patients was more close to the theoretical values of random network at resting state (− 200–0 ms). In Stam's study, similar result was found in AD functional networks (Stam et al., 2007a, 2007b). The large-scale functional brain network organization deviates from the optimal ‘small-world’ network structure towards a more ‘random’ type. This randomized cortical network may reflect regional abnormality and functional disconnection in MCI patients (Liu et al., 2014). Interestingly, we also observed that the properties of cortical networks of MCI patients during the selective attention task were similar to that at resting state. The patients had smaller L and a randomized cortical network in 0–600 ms after the onset of second stimuli. This abnormity may be related with the deficits of selective attention network (Fernandez-Duque and Black, 2008; Krinsky-McHale et al., 2008). The impaired early dysfunction of fronto-parietal networks in MCI patients could result from the decrease of L. Furthermore, in 600– 800 ms, we found that the L of brain network in mismatching task

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

was shorter than in matching task for normal controls and the C of brain network in mismatching task was larger than in matching task for MCI patients. This change of information integration–segregation in MCI patients on late cognitive stage may reflect a true deficit in the organization of attention network in this disorder. In addition, the abnormal properties of cortical networks in MCI patients were both found in alpha and beta bands in this paper. Both of studies on spontaneous and induced brain oscillations demonstrated the MCI patients had decreased alpha and beta synchronization (Basar et al., 2010; Koenig et al., 2005; Park et al., 2008). The decrease may reflect the abnormality of hub regions in the frontal lobe and the temporal lobe (Yao et al., 2010), hippocampus atrophy in MCI patients (Moretti et al., 2007) and the atrophic change in the frontal cortex (Chang et al., 2010). In an attention task, the deficits of cortical upper-alpha activation may represent the attention-executive network impairment in MCI patients (Babiloni et al., 2015). We speculate the functional disconnection and abnormality of whole-brain cortical networks may be due to the deficit of attention network accounting for impaired function of local regions. 4.2. Loss of small-world characteristic in patients with MCI in selective attention task Some recent studies demonstrated that small-world properties are exhibited in functional brain networks and structural brain networks. Compared with random networks, small-world networks have higher clustering coefficients (γ N 1) and similar shortest absolute path length (λ ≈ 1) (Micheloyannis et al., 2006; Stam and Reijneveld, 2007). In summary, the scalar σ = γ/λ will be greater than unity for a smallworld network (Humphries et al., 2006). In present study, the σ of healthy elder networks was larger than one over an entire range of K at resting state. These findings are consistent with those of previous studies that found the human brain to be an efficient neural architecture (Bullmore and Sporns, 2009). Compared with normal controls, lower σ of MCI patients had suggested that the patients had a further loss of small-world attribute in alpha band at resting state. Local and global functional connectivity disruptions in MCI have been found using graph analysis on structural MRI data (Yao et al., 2010). The graph theoretical analysis of positron emission tomography (PET) data revealed the decrease of σ in MCI patients which may represent disrupted information processing among distant brain regions across the whole brain. While, no significant difference for σ was found between MCI and controls (Seo et al., 2013). Our results of EEG at resting state maybe the more effective proof to demonstrate the abnormality of cortical network structure and loss of small-world characteristics in subjects with MCI. In addition, the loss of small-world characteristics of MCI network was found in the early stage of information processing (0–600 ms). It may be the other expression of the attention network deficits in these patients (Deiber et al., 2009). Moreover, in 600–800 ms after the onset of second stimuli, the small-world scalar of MCI patients was the same as that of normal controls. We can speculate that small-world characteristics of MCI network in the postprocessing of conflict information would be normalized. These results may reflect an active compensation mechanism employed to maintain adequate performance for subjects (Geerligs et al., 2012). 4.3. The time-evolution characteristic of cortical network during the color selective attention task In this study, the Time effect was found in the characteristic of cortical networks during a cognition task. In selective attention, alpha band had a central role for attention suppression mechanism when objects need to be specifically ignored or selected against, while beta rhythms can be associated with top-down attention (Foxe and Snyder, 2011; Lee et al., 2013). In this paper, the C of the network in alpha band decreased in 400–800 ms compared with that in resting state and 200–

7

400 ms after stimuli onset. This decline may be associated with the decrease of alpha synchronization oscillation and the reduction of brain interconnection in cognition task. In beta band, the C and L band decreased in 0–200 ms and then increased in 400–600 ms. The increase of C and L showed that the human brain network tends to be a more regular and efficient network in the time course of conflict information processing. Furthermore, we found that the σ in 200–400 ms was different from that in −200–0 ms and 600–800 ms. The special characteristic of brain network in this time range may reflect the detecting and processing of conflict information. At the late stage of information processing, the brain network return to a small world network. In conclusion, we found that the network has a specific network attribute in different time course. On the one hand, the difference of time-evolution characteristic of cortical network indicates that the oscillation of alpha and beta rhythm have their specific role in selective attention task. On the other hand, the change of the network attributes also shows the perception and process of conflict information. 4.4. Mismatch stimulus evoked more regular and efficient brain network Related event-related potentials (ERPs) found conflict information evoked a related negative component N270 which represents the activity of the brain for detecting cognitive information mismatches or conflicts. And high-conflict task can evoke another negative component (N430) which reflects the processing for complex conflicts (Wang et al., 2004; Zhang et al., 2003). In the present study, we found the characteristic of brain network can also reflect the sequential processing of conflict information. The shorter L of network for mismatch stimulus during 200–600 ms means the conflict information needs more strong local interconnectivity among regions. In addition, in 600–800 ms, late conflict processing, we found that the C of network in mismatching task was higher than that in matching task. In other words, the processing of conflict information need stronger information integration and transfer among cortical areas. And the brain network was more efficient during this period of time. Larger σ for mismatch stimulus in beta band showed the brain function network had small-world characteristic. While similar result in alpha band was only found between 200 and 400 ms after stimulus onset. It also reflects that alpha and beta bands had different oscillation characteristics in visual selective attention task (Foxe and Snyder, 2011; Lee et al., 2013). 5. Conclusion In conclusion, the current investigation adopted graph-based analysis to study the whole brain network manifestation of neurological dysfunctions in MCI patients compared with normal controls. The present study suggested that the pathological networks in MCI patients had topological abnormalities in whole brain networks and loss of small-world attributes both during active and resting states. All the characteristics observed in MCI may reflect anatomical structural abnormalities. Such a relationship may contribute to an understanding of the cerebral organization in MCI. Furthermore, we found visual stimuli especially conflict stimuli evoked more efficient brain network. Of course, there are some limitations in our approach. First, only small amounts of EEG data from MCI patients are available for they cannot perform the selective attention task normally. This study included as many subjects as possible to ascertain the real cortical networks accurately. Furthermore, this study is still at its initial stage. We only analyzed the brain network in alpha and beta bands which were correlated with cognitive decline. Future studies in other frequency bands (theta and gamma) should be done to further investigate the abnormal brain network in MCI. In this paper, we only studied the global network manifestation of brain function in MCI and normal controls based on EEG. The integration of multi-level network features obtained with various

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

8

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx

brain imaging technologies will be helpful to understand the pathophysiological mechanism of MCI.

Acknowledgment This study was supported by the National Natural Science Foundation of China (61171032), Open Project of Xuanwu Hospital Capital Medical University-Beijing Key Laboratory of regulate and treat functional brain disease (2013NBTK02).

References Achard, S., Bullmore, E., 2007. Efficiency and cost of economical brain functional networks. PLoS Comput. Biol. 3, e17. Ahmadlou, M., Adeli, A., Bajo, R., Adeli, H., 2014. Complexity of functional connectivity networks in mild cognitive impairment subjects during a working memory task. Clin. Neurophysiol. 125, 694–702. Akay, M., 1995. Introduction: wavelet transforms in biomedical engineering. Ann. Biomed. Eng. 23, 529–530. Albert, M.S., DeKosky, S.T., Dickson, D., 2011. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging—Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 7, 270–279. Babiloni, C., Del, P.C., Boccardi, M., Lizio, R., Lopez, S., Carducci, F., Marzano, N., Soricelli, A., Ferri, R., Triggiani, A.I., Prestia, A., Salinari, S., Rasser, P.E., Basar, E., Fama, F., Nobili, F., Yener, G., Emek-Savas, D.D., Gesualdo, L., Mundi, C., Thompson, P.M., Rossini, P.M., Frisoni, G.B., 2015. Occipital sources of resting-state alpha rhythms are related to local gray matter density in subjects with amnesic mild cognitive impairment and Alzheimer's disease. Neurobiol. Aging 36, 556–570. Baddeley, A.D., Baddeley, H.A., Bucks, R.S., Wilcock, G.K., 2001. Attentional control in Alzheimer's disease. Brain 124, 1492–1508. Bai, F., Zhang, Z., Yu, H., Shi, Y., Yuan, Y., Zhu, W., Zhang, X., Qian, Y., 2008. Default-mode network activity distinguishes amnestic type mild cognitive impairment from healthy aging: a combined structural and resting-state functional MRI study. Neurosci. Lett. 438, 111–115. Basar, E., Guntekin, B., Tulay, E., Yener, G.G., 2010. Evoked and event related coherence of Alzheimer patients manifest differentiation of sensory–cognitive networks. Brain Res. 1357, 79–90. Bassett, D.S., Bullmore, E., 2006. Small-world brain networks. Neuroscientist 12, 512–523. Belleville, S., Chertkow, H., Gauthier, S., 2007. Working memory and control of attention in persons with Alzheimer's disease and mild cognitive impairment. Neuropsychology 21, 458–469. Bian, Z., Li, Q., Wang, L., Lu, C., Yin, S., Li, X., 2014. Relative power and coherence of EEG series are related to amnestic mild cognitive impairment in diabetes. Front. Aging Neurosci. 6, 11. Booth, J.R., Burman, D.D., Meyer, J.R., Lei, Z., Trommer, B.L., Davenport, N.D., Li, W., Parrish, T.B., Gitelman, D.R., Mesulam, M.M., 2005. Larger deficits in brain networks for response inhibition than for visual selective attention in attention deficit hyperactivity disorder (ADHD). J. Child Psychol. Psychiatry 46, 94–111. Bressler, S.L., Menon, V., 2010. Large-scale brain networks in cognition: emerging methods and principles. Trends Cogn. Sci. 14, 277–290. Bressler, S.L., Tang, W., Sylvester, C.M., Shulman, G.L., Corbetta, M., 2008. Top-down control of human visual cortex by frontal and parietal cortex in anticipatory visual spatial attention. J. Neurosci. 28, 10056–10061. Bullmore, E., Sporns, O., 2009. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat. Rev. Neurosci. 10, 186–198. Chang, Y.L., Jacobson, M.W., Fennema-Notestine Jr., C., Hagler, D.J., Jennings, R.G., Dale, A.M., McEvoy, L.K., 2010. Level of executive function influences verbal memory in amnestic mild cognitive impairment and predicts prefrontal and posterior cingulate thickness. Cereb. Cortex 20, 1305–1313. Chen, Z.J., He, Y., Rosa-Neto, P., Germann, J., Evans, A.C., 2008. Revealing modular architecture of human brain structural networks by using cortical thickness from MRI. Cereb. Cortex 18, 2374–2381. Deiber, M.P., Ibanez, V., Missonnier, P., Herrmann, F., Fazio-Costa, L., Gold, G., Giannakopoulos, P., 2009. Abnormal-induced theta activity supports early directed-attention network deficits in progressive MCI. Neurobiol. Aging 30, 1444–1452. Deuker, L., Bullmore, E.T., Smith, M., Christensen, S., Nathan, P.J., Rockstroh, B., Bassett, D.S., 2009. Reproducibility of graph metrics of human brain functional networks. Neuroimage 47, 1460–1468. Fernandez-Duque, D., Black, S.E., 2006. Attentional networks in normal aging and Alzheimer's disease. Neuropsychology 20, 133–143. Fernandez-Duque, D., Black, S.E., 2008. Selective attention in early dementia of Alzheimer type. Brain Cogn. 66, 221–231. Foxe, J.J., Snyder, A.C., 2011. The role of alpha-band brain oscillations as a sensory suppression mechanism during selective attention. Front. Psychol. 2, 154. Gauthier, S., Reisberg, B., Zaudig, M., Petersen, R.C., Ritchie, K., Broich, K., Belleville, S., Brodaty, H., Bennett, D., Chertkow, H., Cummings, J.L., de Leon, M., Feldman, H., Ganguli, M., Hampel, H., Scheltens, P., Tierney, M.C., Whitehouse, P., Winblad, B., 2006. Mild cognitive impairment. Lancet 367, 1262–1270. Gazzaley, A., Nobre, A.C., 2012. Top-down modulation: bridging selective attention and working memory. Trends Cogn. Sci. 16, 129–135.

Geerligs, L., Saliasi, E., Maurits, N.M., Lorist, M.M., 2012. Compensation through increased functional connectivity: neural correlates of inhibition in old and young. J. Cogn. Neurosci. 24, 2057–2069. Geerligs, L., Saliasi, E., Maurits, N.M., Renken, R.J., Lorist, M.M., 2014. Brain mechanisms underlying the effects of aging on different aspects of selective attention. Neuroimage 91, 52–62. Greicius, M.D., Srivastava, G., Reiss, A.L., Menon, V., 2004. Default-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional MRI. Proc. Natl. Acad. Sci. U. S. A. 101, 4637–4642. He, Y., Chen, Z.J., Evans, A.C., 2007. Small-world anatomical networks in the human brain revealed by cortical thickness from MRI. Cereb. Cortex 17, 2407–2419. Humphries, M.D., Gurney, K., Prescott, T.J., 2006. The brainstem reticular formation is a small-world, not scale-free, network. Proc. Biol. Sci. 273, 503–511. Jeong, J., 2004. EEG dynamics in patients with Alzheimer's disease. Clin. Neurophysiol. 115, 1490–1505. Jiang, X., Zhu, D., Li, K., Zhang, T., Wang, L., Shen, D., Guo, L., Liu, T., 2014. Predictive models of resting state networks for assessment of altered functional connectivity in mild cognitive impairment. Brain Imaging Behav. 8 (4), 542–557. Knyazeva, M.G., Jalili, M., Brioschi, A., Bourquin, I., Fornari, E., Hasler, M., Meuli, R., Maeder, P., Ghika, J., 2010. Topography of EEG multivariate phase synchronization in early Alzheimer's disease. Neurobiol. Aging 31, 1132–1144. Ko, P.C., Kilduff, P.T., Higgens, J.A., Milberg, W., McGlinchey, R., 2005. Evidence for intact selective attention in Alzheimer's disease patients using a location priming task. Neuropsychology 19, 381–389. Koenig, T., Prichep, L., Dierks, T., Hubl, D., Wahlund, L.O., John, E.R., Jelic, V., 2005. Decreased EEG synchronization in Alzheimer's disease and mild cognitive impairment. Neurobiol. Aging 26, 165–171. Krinsky-McHale, S.J., Devenny, D.A., Kittler, P., Silverman, W., 2008. Selective attention deficits associated with mild cognitive impairment and early stage Alzheimer's disease in adults with Down syndrome. Am. J. Ment. Retard. 113, 369–386. Lee, J.H., Whittington, M.A., Kopell, N.J., 2013. Top-down beta rhythms support selective attention via interlaminar interaction: a model. PLoS Comput. Biol. 9, e1003164. Liu, Z., Zhang, Y., Yan, H., Bai, L., Dai, R., Wei, W., Zhong, C., Xue, T., Wang, H., Feng, Y., You, Y., Zhang, X., Tian, J., 2012. Altered topological patterns of brain networks in mild cognitive impairment and Alzheimer's disease: a resting-state fMRI study. Psychiatry Res. 202, 118–125. Liu, Y., Yu, C., Zhang, X., Liu, J., Duan, Y., Alexander-Bloch, A.F., Liu, B., Jiang, T., Bullmore, E., 2014. Impaired long distance functional connectivity and weighted network architecture in Alzheimer's disease. Cereb. Cortex 24, 1422–1435. Micheloyannis, S., Pachou, E., Stam, C.J., Breakspear, M., Bitsios, P., Vourkas, M., Erimaki, S., Zervakis, M., 2006. Small-world networks and disturbed functional connectivity in schizophrenia. Schizophr. Res. 87, 60–66. Moretti, D.V., Miniussi, C., Frisoni, G.B., Geroldi, C., Zanetti, O., Binetti, G., Rossini, P.M., 2007. Hippocampal atrophy and EEG markers in subjects with mild cognitive impairment. Clin. Neurophysiol. 118, 2716–2729. Park, Y.M., Che, H.J., Im, C.H., Jung, H.T., Bae, S.M., Lee, S.H., 2008. Decreased EEG synchronization and its correlation with symptom severity in Alzheimer's disease. Neurosci. Res. 62, 112–117. Petersen, R.C., 2004. Mild cognitive impairment as a diagnostic entity. J. Intern. Med. 256, 183–194. Pignatti, R., Rabuffetti, M., Imbornone, E., Mantovani, F., Alberoni, M., Farina, E., Canal, N., 2005. Specific impairments of selective attention in mild Alzheimer's disease. J. Clin. Exp. Neuropsychol. 27, 436–448. Ponten, S.C., Bartolomei, F., Stam, C.J., 2007. Small-world networks and epilepsy: graph theoretical analysis of intracerebrally recorded mesial temporal lobe seizures. Clin. Neurophysiol. 118, 918–927. Portet, F., Ousset, P.J., Visser, P.J., Frisoni, G.B., Nobili, F., Scheltens, P., Vellas, B., Touchon, J., 2006. Mild cognitive impairment (MCI) in medical practice: a critical review of the concept and new diagnostic procedure. Report of the MCI Working Group of the European Consortium on Alzheimer's disease. J. Neurol. Neurosurg. Psychiatry 77, 714–718. Quian, Q.R., Kraskov, A., Kreuz, T., Grassberger, P., 2002. Performance of different synchronization measures in real data: a case study on electroencephalographic signals. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 65, 041903. Rombouts, S.A., Barkhof, F., Goekoop, R., Stam, C.J., Scheltens, P., 2005. Altered resting state networks in mild cognitive impairment and mild Alzheimer's disease: an fMRI study. Hum. Brain Mapp. 26, 231–239. Rosenblum, M.G., Pikovsky, A.S., Kurths, J., 1996. Phase synchronization of chaotic oscillators. Phys. Rev. Lett. 76, 1804–1807. Seo, E.H., Lee, D.Y., Lee, J.M., Park, J.S., Sohn, B.K., Lee, D.S., Choe, Y.M., Woo, J.I., 2013. Whole-brain functional networks in cognitively normal, mild cognitive impairment, and Alzheimer's disease. PLoS One 8, e53922. Sporns, O., Zwi, J.D., 2004. The small world of the cerebral cortex. Neuroinformatics 2, 145–162. Stam, C.J., Reijneveld, J.C., 2007. Graph theoretical analysis of complex networks in the brain. Nonlinear Biomed. Phys. 1, 3. Stam, C.J., van Straaten, E.C., 2012. The organization of physiological brain networks. Clin. Neurophysiol. 123, 1067–1087. Stam, C.J., Jones, B.F., Nolte, G., Breakspear, M., Scheltens, P., 2007a. Small-world networks and functional connectivity in Alzheimer's disease. Cereb. Cortex 17, 92–99. Stam, C.J., Nolte, G., Daffertshofer, A., 2007b. Phase lag index: assessment of functional connectivity from multi channel EEG and MEG with diminished bias from common sources. Hum. Brain Mapp. 28, 1178–1193. Sun, J., Small, M., 2009. Unified framework for detecting phase synchronization in coupled time series. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 80, 046219.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015

L. Wei et al. / International Journal of Psychophysiology xxx (2015) xxx–xxx Uhlhaas, P.J., Singer, W., 2006. Neural synchrony in brain disorders: relevance for cognitive dysfunctions and pathophysiology. Neuron 52, 155–168. Wang, Y., Cui, L., Wang, H., Tian, S., Zhang, X., 2004. The sequential processing of visual feature conjunction mismatches in the human brain. Psychophysiology 41, 21–29. Wang, L., Guo, X., Sun, J., Jin, Z., Tong, S., 2012. Cortical networks of hemianopia stroke patients: a graph theoretical analysis of EEG signals at resting state. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2012, 49–52. Watts, D.J., Strogatz, S.H., 1998. Collective dynamics of ‘small-world’ networks. Nature 393, 440–442. Winblad, B., Palmer, K., Kivipelto, M., Jelic, V., Fratiglioni, L., Wahlund, L.O., Nordberg, A., Backman, L., Albert, M., Almkvist, O., Arai, H., Basun, H., Blennow, K., de Leon, M., DeCarli, C., Erkinjuntti, T., Giacobini, E., Graff, C., Hardy, J., Jack, C., Jorm, A., Ritchie, K., van Duijn, C., Visser, P., Petersen, R.C., 2004. Mild cognitive impairment—beyond

9

controversies, towards a consensus: report of the International Working Group on Mild Cognitive Impairment. J. Intern. Med. 256, 240–246. Wu, J., Zhang, J., Liu, C., Liu, D., Ding, X., Zhou, C., 2012. Graph theoretical analysis of EEG functional connectivity during music perception. Brain Res. 1483, 71–81. Yao, Z., Zhang, Y., Lin, L., Zhou, Y., Xu, C., Jiang, T., 2010. Abnormal cortical networks in mild cognitive impairment and Alzheimer's disease. PLoS Comput. Biol. 6, e1001006. Zhang, X., Wang, Y., Li, S., Wang, L., 2003. Event-related potential N270, a negative component to identification of conflicting information following memory retrieval. Clin. Neurophysiol. 114, 2461–2468. Zhang, S., Xu, M., Kamigaki, T., Hoang, D.J., Chang, W.C., Jenvay, S., Miyamichi, K., Luo, L., Dan, Y., 2014. Selective attention. Long-range and local circuits for top-down modulation of visual cortex processing. Science 345, 660–665.

Please cite this article as: Wei, L., et al., Altered characteristic of brain networks in mild cognitive impairment during a selective attention task: An EEG study, Int. J. Psychophysiol. (2015), http://dx.doi.org/10.1016/j.ijpsycho.2015.05.015