Accepted Manuscript Disturbed phase relations in white matter hyperintensity based vascular dementia: An EEG directed connectivity study E.C.W. van Straaten, J. den Haan, H. de Waal, W.M. van der Flier, F. Barkhof, N.D. Prins, C.J. Stam PII: DOI: Reference:
S1388-2457(14)00293-4 http://dx.doi.org/10.1016/j.clinph.2014.05.018 CLINPH 2007115
To appear in:
Clinical Neurophysiology
Accepted Date:
21 May 2014
Please cite this article as: van Straaten, E.C.W., den Haan, J., de Waal, H., van der Flier, W.M., Barkhof, F., Prins, N.D., Stam, C.J., Disturbed phase relations in white matter hyperintensity based vascular dementia: An EEG directed connectivity study, Clinical Neurophysiology (2014), doi: http://dx.doi.org/10.1016/j.clinph.2014.05.018
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
dPLI in WMH – van Straaten et al.
Disturbed phase relations in white matter hyperintensity based vascular dementia: An EEG directed connectivity study E.C.W. van Straaten1,4* J. den Haan1, H. de Waal2, W.M. van der Flier2,3, F. Barkhof4, N.D. Prins2, C.J. Stam1 1
Department of Clinical Neurophysiology, 2 Alzheimer Center at department of Neurology,
3
Department of Epidemiology and biostatistics, and 4Department of Radiology; Neuroscience
Campus Amsterdam, VU University Medical Center, Amsterdam, The Netherlands. 4Nutricia Advanced Medical Nutrition,Danone Research, Centre for Specialised Nutrition, Wageningen, The Netherlands.
Corresponding author: Department of Clinical Neurophysiology, VU University Medical Centre, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands Tel: +31 20 444 0730 Fax: +31 20 444 4816 E-mail: [email protected]
Highlights -
Directed functional connectivity shows a phase lead of the frontal regions to parietooccipital regions in subjects without objective cognitive disturbance.
-
In white matter hyperintensity (WMH) based vascular dementia (VaD) this organization of phase relationships is lost.
-
Undirected functional connectivity does not differ between control subjects and WMHVaD patients.
dPLI in WMH – van Straaten et al. Abstract
Objective White matter hyperintensities (WMH), a feature seen on magnetic resonance imaging (MRI) and regarded to reflect small vessel disease, can lead to vascular dementia (WMHVaD). In WMH-VaD, cognitive deficits typically consist of executive function disturbances, and reduced information processing speed, regarded as a result of cerebral hypoperfusion. We aimed to investigate whether this patient group has typical functional differences from controls. Methods Resting-state encephalography studies of 17 VaD patients and 17 age- and gender matched non-demented controls were analyzed in the delta, theta, alpha1 and 2, and beta frequency bands. Undirected functional connectivity between electrodes was established with the Phase Lag Index (PLI) and directed functional connectivity with the directed phase lag index (dPLI). PLI and dPLI were related to performance in cognitive testing. Results Mean PLI did not differ between patients and controls. In the control group dPLI showed anterior to posterior phase gradients in all bands except the delta band. In the VaD patient group this pattern was significantly different without a clear directional pattern. No relationship with cognition was demonstrated. Conclusions This study shows a clear front-to-back direction of connectivity in non-demented controls. In VaD patients with extensive WMH, this pattern is disturbed. Significance Structural damage at the regions of long distance white matter tracts may induce changes in the direction of phase relationships of distinct brain regions.
Keywords: Phase Lag Index, directed phase lag index, vascular dementia, WMH, functional connectivity, EEG.
dPLI in WMH – van Straaten et al. Introduction White matter hyperintensities White matter hyperintensities (WMH) are a frequent finding on magnetic resonance imaging (MRI) of elderly subjects, mainly as punctiform lesions (Breteler et al., 1994;Longstreth, Jr. et al., 1996). However, WMH can also increase to large, confluent areas and lead to dementia (WMH-VaD). The cognitive profile of subjects WMH-VaD is classically dominated by executive function disturbances and reduced information processing speed (Au et al., 2006;Kramer et al., 2002;Prins et al., 2005). The mechanisms underlying these cognitive deficits remain largely unknown. The typical locations of WMH are the periventricular regions and deep white matter sparing the u-fibres. These regions harbour the long white matter tracts that connect longer distant brain regions, for example the bidirectional superior longitudinal fascicle connecting frontal and posterior brain areas. We hypothesize that the cognitive deficits seen in WMH-based dementia might be caused by dysfunction of these white matter tracts and that this dysfunction is reflected by disturbance of functional connectivity of brain regions on neurophysiological testing.
EEG in VaD studies Electroencephalography (EEG) can be used to study the effect of vascular lesions on brain electrical activity. Few EEG studies on WMH have been published (Erkinjuntti et al., 1988;Gawel et al., 2007;Liedorp et al., 2009;Signorino et al., 1996). In these studies, WMH were studied mostly together with other vascular lesions, such as lacunes and cortical infarcts. Slowing of the background rhythm was the main finding on quantitative analysis. In a former study based on the same patient sample, we showed that the slowing of the resting-state oscillations was severe in WMH-VaD (van Straaten et al., 2012). However, this finding is nonspecific, and does not in itself provide information on the pathophysiological mechanism in VaD. The analysis of directed and undirected functional connectivity might aid in entangling the mode of action by which WMH exert their action on cognition.
Functional connectivity and WMH Synchronization of neuronal activity between brain regions is reflected by correlations in oscillatory activity in EEG, magnetic encephalography (MEG), and functional magnetic resonance imaging (fMRI) studies. For WMH, a few activation studies investigating the effect on brain activity as measured with fMRI have been reported. Linortner et al. found an increase in activated brain area during a simple motor task in subjects with a high WMH
dPLI in WMH – van Straaten et al. load compared to subjects with no or only mild WMH (Linortner et al., 2012). This result was interpreted as possibly due to a compensatory mechanism. Nordahl et al. and Venkatraman et al. reported a decrease in activated brain area during memory and executive tasks respectively (Nordahl et al., 2006;Venkatraman et al., 2010). One study mentioned the investigation of the undirected functional connectivity within the default mode network, a cluster of functionally coupled brain regions that is active when the subject is in an awake, eyes-closed resting state. The investigators found no relationship between WMH volume and connectivity (He et al., 2012).
Direction and delays of information processing In connectivity studies the direction of the interactions can be taken into account. In a recent paper the direction of flow, defined as the extent to which one region is leading or lagging in phase compared to all other regions, was reported to have a front-to-back orientation in an anatomically realistic computational model of coupled neural masses (Stam and van Straaten, 2012). This is in agreement with previous EEG and MEG studies that have found an anterior to posterior flow in slow and fast oscillatory brain activity, slow activity in sleep, and the EEG phenomenon of triphasic waves (Boulanger et al., 2006;Llinas and Ribary, 1993;Massimini et al., 2004;Rana et al., 2012). In addition, Leech et al. and Yan and He reported posterior parts of fMRI resting-state brain networks to be driven by other parts of the networks (Leech et al., 2012;Yan and He, 2011). On the other hand, two studies of one research group reported an opposite, parietal to frontal, direction of flow in resting-state EEGs of healthy elderly (Babiloni et al., 2008;Babiloni et al., 2009). It can be conceived that the exact timing of the neuronal firing in functionally coupled regions is disturbed by dysfunction of the anatomical connections between them, and that this will be expressed as a disturbance and dispersion of the directional pattern. A consistent phase lag between two time series, such as EEG electrodes, is indicative of a reproducible, stable and thereby intact neuronal conductivity. A lack of consistent phase lag can be seen when no functional connection is present, or when the conductivity is non uniformly delayed and dispersed. We hypothesize that this may be the case in severe WMH.
Aim of the study We investigate whether WMH-based VaD is mainly characterized by loss of global connectivity with preserved directionality or that the direction of the coupling between
dPLI in WMH – van Straaten et al. regions is impaired. We compare WMH-based VaD patients to age and gender matched controls with respect to their EEG-derived functional brain connectivity.
Methods Participants Subjects were selected from the memory clinic based Amsterdam Dementia Cohort. They had been referred to the Alzheimer Center of the VU University Medical Center between 2003 and January 2010 and had undergone diagnostic workup including neuropsychological testing, brain MRI (or, if not possible, computed tomography (CT)), and resting-state EEG. All subjects gave informed consent for the storage of the results of the examinations in a local database and for the use of the data for research purposes. This protocol was in agreement with the declaration of Helsinki and approved by the ethical review board of the VU University Medical Center. During a multidisciplinary consensus meeting, patients were diagnosed by consensus. The NINDS-AIREN criteria were applied for the diagnosis probable VaD based on WMH (dementia syndrome and WMH involving at least 25% of the total white matter) (Roman et al., 1993). Individuals with other causes for VaD (cortical infarctions or multiple lacunes in strategic areas) were not included, as were subjects with other MRI abnormalities possibly interfering with the diagnosis. Additionally, the subjects did not meet the criteria for Alzheimer’s disease (AD) (McKhann et al., 1984). As a result of this selection procedure, all subjects had a severe global WMH score and no other suspected cause for the dementia syndrome. The controls were individuals from the same database who underwent the screening because of subjective cognitive complaints but who had no cognitive abnormalities on testing. Subjects were individually matched (one control for each VaD patient) for age and gender to the patients. The study groups were identical to the dataset used in a previous study on quantitative EEG (van Straaten et al., 2012). Table 1 shows demographic characteristics of the two groups.
EEG recordings At the first visit resting state EEG of 20 minutes was recorded in all subjects with OSG digital equipment (Brainlab®; OSG b.v. Belgium). Twenty-one electrodes were placed according to the 10–20 international system (Fp2, Fp1, F8, F7, F4, F3, A2, A1 T4, T3, C4, C3, T6, T5, P4, P3, O2, O1, Fz, Pz, and Cz). Electrode impedance was below 5 kΩ. Filter settings were: time constant 1 s, and low pass filter 70 Hz. Sample frequency was 500 Hz and analog–digital precision was 16 bit. EEGs were recorded (all electrodes separately and
dPLI in WMH – van Straaten et al. using an average reference). Subjects were seated in a slightly reclined chair in a soundattenuated, normally lit room and were kept awake as much as possible during the recording. For each subject, four epochs (average reference montage) of 10 seconds (5000 samples) each were visually selected on the basis of signal quality (artefact-free) and patient status (eyes closed, awake), then converted to ASCII format. The first 4096 samples of each epoch were used for further analysis (BrainWave analysis software, version 0.9.72, Stam, 2012; available at: http:/home.kpn.nl/stam7883/brainwave.html). EEGs were digitally rereferenced offline to the average A1/A2 reference derivation. Subsequently, a Hilbert transform was used to compare signals.
Functional connectivity analysis For each EEG epoch, we calculated the undirected functional connectivity with the Phase Lag Index (PLI). With this method, the distribution of differences of the instantaneous phases, determined from a Hilbert transform, between every channel pair is computed, resulting in a PLI value between 0 and 1 (0 ≤ PLI ≤ 1) for each channel pair:
PLI = sign[∆φ (tk )]
(1)
where ∆φ (tk ) , k = 1 . . . N is a time series of the phase differences.
Higher PLI values mean higher consistency of coupling, a value of 0 indicates no coupling or coupling with a phase difference of zero (which cannot reliably be distinguished from volume conduction effects) (Stam et al., 2007). A PLI value > 0 indicates a true interaction without noise from volume conduction and without bias from an active reference in the case of EEG. PLI values were calculated in five frequency bands (delta 0.5–4 Hz, theta 4–8 Hz, alpha1 8–10 Hz, alpha2 10-13 Hz, and beta 13–30 Hz) in an average reference derivation (using all electrodes except FP2/1 and A2/1). We left gamma band (> 30 Hz) activity out of the analyses because this fast activity cannot reliably be distinguished from muscle artefact (Whitham et al., 2007). PLI per subject was calculated by taking the mean of the four epochs. This was done for all EEG channels separately, as well as for the mean of all channels. For regional differences we clustered the channels to render six regions for further analysis, as used in a previous study: frontocentral right (Fp2, F4, C4), frontocentral left (Fp1, F3, C3), temporal right (F8, T4, T6), temporal left (F7, T3, T5), posterior right (P4, O2), and posterior left (P3, O1).
Measurement of direction and delay
dPLI in WMH – van Straaten et al.
In addition to the PLI, we computed for all the EEG epochs the directed Phase Lag Index (dPLI). This is a recently introduced connectivity measure that not only establishes the distribution of the instantaneous phase differences between channels (in the same manner as with the PLI), but also indicates what the direction of the differences is (Stam and van Straaten, 2012). To put differently: of every EEG electrode channel pair, in every time point, it is established which channel is leading and which channel is lagging in phase and how much by:
dPLIxy =
1 N
N
∑ H (φ ) t
(2)
t =1
where H is the Heaviside step function and
φt , t = 1 .... N a time series of phase differences
of time series x and y. Directed PLI ranges from 0 – 1 (0 ≤ dPLI≤ 1). For each channel pair xy, differences of the instantaneous phases, again determined from a Hilbert transform, in each time point are computed. The corresponding dPLI value is based on the direction of this difference, and a mean dPLI was computed for all channels separately by taking the mean of the four epochs. When no consisting phase leading or lagging is measured, the two channels are each leading in an equal amount. This will give a dPLI value of 0.5 by convention. When one channel is constantly leading compared to the other, this will result in a dPLI value for the leading channel of 1 and a dPLI value of almost 0 for the channel that is lagging. For illustration, Figure 1 shows the output of a dPLI computation of one epoch of a control subject. The dPLI cannot be averaged over all EEG channels, since this will result in a value of 0.5: the leading channels average out the lagging channels. For the computation and assessment of the differences between groups, we clustered the channels into the same six regions as described for the undirected connectivity: bilaterally frontocentral, temporal, and posterior.
Relationship with cognitive impairment In the patient group, connectivity (undirected and directed) was related to neuropsychological tests to assess the association between functional connectivity and cognitive performance. As a global cognitive measure we used the Mini Mental State Examination (MMSE) and for the assessment of memory function we used the Rey auditory verbal learning test (RAVLT, sum of five learning trials for immediate recall and a delayed recall trial after 20 minutes) (Folstein et al., 1975;Rey, 1958). Trailmaking test B (TMT B) and verbal fluency (animals) test were used as measures for executive functioning
dPLI in WMH – van Straaten et al.
(Reitan and Wolfson, 2004;Spreen and Straus, 1998). Missing values due to the subject’s cognitive incapability to perform or complete the test were set at the lowest possible score for that test.
Statistical analysis The distribution of functional connectivity values was checked with the KolmogorovSmirnov test. Small deviations from a normal distribution were accepted. Paired t-tests were used for comparison of mean PLI. No mean dPLI was tested since it follows from the method that the mean dPLI over all channels is 0.5. To test for group and regional differences in dPLI, a repeated-measures analysis of variance (ANOVA) with GreenhouseGeisser correction for sphericity was performed, with group as the between subject variable and the region (frontocentral, temporal, and posterior; right and left) as the within subject variable. In frequency bands with a significant difference between groups or with a significant interaction between group and area in the ANOVA, a post-hoc permutation test was performed to display the brain areas with significant differences between the groups: Directed PLI differences between the groups were compared to differences of two groups with randomly assigned subjects (repeated 100 times). In the patient group the EEG measures per frequency band were related to the cognitive measures using linear regression analysis with PLI and dPLI (of region frontocentral right) as independent variables and MMSE, RAVLT, TMT B and verbal fluency as the dependent variables.
Results
General characteristics and functional connectivity In total, seventeen WMH-VaD patients were identified meeting inclusion criteria. Three individuals in the patient group and two in the control group had CT available instead of MRI. PLI values of the WMH-VaD patients did not differ significantly from those of the controls in any band (Table 2). Also, no interaction between the regional PLI and group was found (F ≤ 0.80, p > 0.05).
Direction and delay Directed PLI differed significantly between groups. Group differences were found with ANOVA in the beta band (F 7.71, p 0.5 (F3 leading over O1) and all lines under the x-axis (negative y-value) correspond to a dPLI value < 0.5 (O1 leading over F3). More lines are located above the xaxis, and this is confirmed by the mean dPLI value of the whole epoch: 0.77. C. Colourcoded dPLI values for all channels for the first 50% of the epoch. X-axis: time. Y-axis: electrodes (numbers correspond with the order in which the electrodes are mentioned in paragraph 2.2). Blue colour refers to a dPLI value < 0.5, red colour indicates dPLI > 0.5. The posterior regions (electrodes 13 – 18, and 21 are relatively more blue (lagging) compared to the anterior regions (electrodes 1 – 12, 19, and 20). D. dPLI anatomically projected (nose up, left = left hemisphere), red is high dPLI, blue is low dPLI. Lines are connections between brain areas, based on the Phase Lag Index (PLI). A front-to-back (from red to blue) distribution of dPLI can be seen. The colour bar applies to both Figure 1C and D.
Figure 2. Topographical representation (nose up, left = left hemisphere) of average directed Phase Lag Index (dPLI) for each group. Red indicates relatively high value, blue indicates relatively low value. Connections between brain regions based on the Phase Lag Index are indicated with lines. Stars indicate brain area with difference between patients and controls (post-hoc permutation analysis, p 0 and dPLI between 0 and 0.5 in the channel that is lagging and dPLI between 0.5 and 1 in the channel that is leading. B: presumed situation in WMH-VaD with phase differences between channels ≠ 0, and with alternatingly leading and lagging of channels. This corresponds to a PLI > 0, and dPLI close to 0.5.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
dPLI in WMH – van Straaten et al.
Table 1. Characteristics of white matter intensities based vascular dementia patients and control subjects. Patients (n=17)
Control subjects (n=17)
Age (yrs )
74.6 ± 8
74.4 ± 8
Sex ratio (m/f)
9/8
9/8
Mean WMH score (SD)
3 ± 0*
1 ± 1*
Mean MMSE score (SD)
22 ± 5*
28.5 ± 1*
Mean MTA score (SD)
1.7 (0.6)
1.0 (0.8)
MMSE: mini mental state examination, MTA: medial temporal lobe atrophy, WMH: white matter hyperintensities. *: p < 0.05 for difference between groups
dPLI in WMH – van Straaten et al.
Table 2. Functional connectivity values for the three subject groups.
Mean PLI control subjects (SD)
Mean PLI VaD patients (SD)
Delta
Theta
Alpha 1
0.15 (0.02)
0.17 (0.07)
0.25 (0.08)
0.15 (0.03)
0.17 (0.04)
0.22 (0.07)
N = 17 in each group. PLI: Phase Lag Index.
Alpha 2
Beta
0.15
0.08
(0.03)
(0.01)
0.16
0.07
(0.03)
(0.03)
dPLI in WMH – van Straaten et al.
Table 3. Analysis of variance (repeated measures) for regional directed Phase Lag Index values in each frequency band. Within subjects Area
Area x Group
Between subjects Group
Delta band
F [3.11]=4.38**
F [3.11]=2.22
F [1]=4.59*
Theta band
F [1.66]=2.13
F [1.66]=2.34
F [1]=4.16*
Alpha 1 band
F [2.00]=11.90**
F [2.00]=3.28*
F [1]=5.36*
Alpha 2 band
F [2.45]=1.52
F [2.45]=4.09*
F [1]=1.68
Beta band
F [2.19]=2.27
F [2.19]=4.82**
F [1]=7.71**
* p < 0.05 ** p < 0.01 Degrees of freedom are printed between square brackets.
S1388-2457(14)00293-4 http://dx.doi.org/10.1016/j.clinph.2014.05.018 CLINPH 2007115
To appear in:
Clinical Neurophysiology
Accepted Date:
21 May 2014
Please cite this article as: van Straaten, E.C.W., den Haan, J., de Waal, H., van der Flier, W.M., Barkhof, F., Prins, N.D., Stam, C.J., Disturbed phase relations in white matter hyperintensity based vascular dementia: An EEG directed connectivity study, Clinical Neurophysiology (2014), doi: http://dx.doi.org/10.1016/j.clinph.2014.05.018
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
dPLI in WMH – van Straaten et al.
Disturbed phase relations in white matter hyperintensity based vascular dementia: An EEG directed connectivity study E.C.W. van Straaten1,4* J. den Haan1, H. de Waal2, W.M. van der Flier2,3, F. Barkhof4, N.D. Prins2, C.J. Stam1 1
Department of Clinical Neurophysiology, 2 Alzheimer Center at department of Neurology,
3
Department of Epidemiology and biostatistics, and 4Department of Radiology; Neuroscience
Campus Amsterdam, VU University Medical Center, Amsterdam, The Netherlands. 4Nutricia Advanced Medical Nutrition,Danone Research, Centre for Specialised Nutrition, Wageningen, The Netherlands.
Corresponding author: Department of Clinical Neurophysiology, VU University Medical Centre, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands Tel: +31 20 444 0730 Fax: +31 20 444 4816 E-mail: [email protected]
Highlights -
Directed functional connectivity shows a phase lead of the frontal regions to parietooccipital regions in subjects without objective cognitive disturbance.
-
In white matter hyperintensity (WMH) based vascular dementia (VaD) this organization of phase relationships is lost.
-
Undirected functional connectivity does not differ between control subjects and WMHVaD patients.
dPLI in WMH – van Straaten et al. Abstract
Objective White matter hyperintensities (WMH), a feature seen on magnetic resonance imaging (MRI) and regarded to reflect small vessel disease, can lead to vascular dementia (WMHVaD). In WMH-VaD, cognitive deficits typically consist of executive function disturbances, and reduced information processing speed, regarded as a result of cerebral hypoperfusion. We aimed to investigate whether this patient group has typical functional differences from controls. Methods Resting-state encephalography studies of 17 VaD patients and 17 age- and gender matched non-demented controls were analyzed in the delta, theta, alpha1 and 2, and beta frequency bands. Undirected functional connectivity between electrodes was established with the Phase Lag Index (PLI) and directed functional connectivity with the directed phase lag index (dPLI). PLI and dPLI were related to performance in cognitive testing. Results Mean PLI did not differ between patients and controls. In the control group dPLI showed anterior to posterior phase gradients in all bands except the delta band. In the VaD patient group this pattern was significantly different without a clear directional pattern. No relationship with cognition was demonstrated. Conclusions This study shows a clear front-to-back direction of connectivity in non-demented controls. In VaD patients with extensive WMH, this pattern is disturbed. Significance Structural damage at the regions of long distance white matter tracts may induce changes in the direction of phase relationships of distinct brain regions.
Keywords: Phase Lag Index, directed phase lag index, vascular dementia, WMH, functional connectivity, EEG.
dPLI in WMH – van Straaten et al. Introduction White matter hyperintensities White matter hyperintensities (WMH) are a frequent finding on magnetic resonance imaging (MRI) of elderly subjects, mainly as punctiform lesions (Breteler et al., 1994;Longstreth, Jr. et al., 1996). However, WMH can also increase to large, confluent areas and lead to dementia (WMH-VaD). The cognitive profile of subjects WMH-VaD is classically dominated by executive function disturbances and reduced information processing speed (Au et al., 2006;Kramer et al., 2002;Prins et al., 2005). The mechanisms underlying these cognitive deficits remain largely unknown. The typical locations of WMH are the periventricular regions and deep white matter sparing the u-fibres. These regions harbour the long white matter tracts that connect longer distant brain regions, for example the bidirectional superior longitudinal fascicle connecting frontal and posterior brain areas. We hypothesize that the cognitive deficits seen in WMH-based dementia might be caused by dysfunction of these white matter tracts and that this dysfunction is reflected by disturbance of functional connectivity of brain regions on neurophysiological testing.
EEG in VaD studies Electroencephalography (EEG) can be used to study the effect of vascular lesions on brain electrical activity. Few EEG studies on WMH have been published (Erkinjuntti et al., 1988;Gawel et al., 2007;Liedorp et al., 2009;Signorino et al., 1996). In these studies, WMH were studied mostly together with other vascular lesions, such as lacunes and cortical infarcts. Slowing of the background rhythm was the main finding on quantitative analysis. In a former study based on the same patient sample, we showed that the slowing of the resting-state oscillations was severe in WMH-VaD (van Straaten et al., 2012). However, this finding is nonspecific, and does not in itself provide information on the pathophysiological mechanism in VaD. The analysis of directed and undirected functional connectivity might aid in entangling the mode of action by which WMH exert their action on cognition.
Functional connectivity and WMH Synchronization of neuronal activity between brain regions is reflected by correlations in oscillatory activity in EEG, magnetic encephalography (MEG), and functional magnetic resonance imaging (fMRI) studies. For WMH, a few activation studies investigating the effect on brain activity as measured with fMRI have been reported. Linortner et al. found an increase in activated brain area during a simple motor task in subjects with a high WMH
dPLI in WMH – van Straaten et al. load compared to subjects with no or only mild WMH (Linortner et al., 2012). This result was interpreted as possibly due to a compensatory mechanism. Nordahl et al. and Venkatraman et al. reported a decrease in activated brain area during memory and executive tasks respectively (Nordahl et al., 2006;Venkatraman et al., 2010). One study mentioned the investigation of the undirected functional connectivity within the default mode network, a cluster of functionally coupled brain regions that is active when the subject is in an awake, eyes-closed resting state. The investigators found no relationship between WMH volume and connectivity (He et al., 2012).
Direction and delays of information processing In connectivity studies the direction of the interactions can be taken into account. In a recent paper the direction of flow, defined as the extent to which one region is leading or lagging in phase compared to all other regions, was reported to have a front-to-back orientation in an anatomically realistic computational model of coupled neural masses (Stam and van Straaten, 2012). This is in agreement with previous EEG and MEG studies that have found an anterior to posterior flow in slow and fast oscillatory brain activity, slow activity in sleep, and the EEG phenomenon of triphasic waves (Boulanger et al., 2006;Llinas and Ribary, 1993;Massimini et al., 2004;Rana et al., 2012). In addition, Leech et al. and Yan and He reported posterior parts of fMRI resting-state brain networks to be driven by other parts of the networks (Leech et al., 2012;Yan and He, 2011). On the other hand, two studies of one research group reported an opposite, parietal to frontal, direction of flow in resting-state EEGs of healthy elderly (Babiloni et al., 2008;Babiloni et al., 2009). It can be conceived that the exact timing of the neuronal firing in functionally coupled regions is disturbed by dysfunction of the anatomical connections between them, and that this will be expressed as a disturbance and dispersion of the directional pattern. A consistent phase lag between two time series, such as EEG electrodes, is indicative of a reproducible, stable and thereby intact neuronal conductivity. A lack of consistent phase lag can be seen when no functional connection is present, or when the conductivity is non uniformly delayed and dispersed. We hypothesize that this may be the case in severe WMH.
Aim of the study We investigate whether WMH-based VaD is mainly characterized by loss of global connectivity with preserved directionality or that the direction of the coupling between
dPLI in WMH – van Straaten et al. regions is impaired. We compare WMH-based VaD patients to age and gender matched controls with respect to their EEG-derived functional brain connectivity.
Methods Participants Subjects were selected from the memory clinic based Amsterdam Dementia Cohort. They had been referred to the Alzheimer Center of the VU University Medical Center between 2003 and January 2010 and had undergone diagnostic workup including neuropsychological testing, brain MRI (or, if not possible, computed tomography (CT)), and resting-state EEG. All subjects gave informed consent for the storage of the results of the examinations in a local database and for the use of the data for research purposes. This protocol was in agreement with the declaration of Helsinki and approved by the ethical review board of the VU University Medical Center. During a multidisciplinary consensus meeting, patients were diagnosed by consensus. The NINDS-AIREN criteria were applied for the diagnosis probable VaD based on WMH (dementia syndrome and WMH involving at least 25% of the total white matter) (Roman et al., 1993). Individuals with other causes for VaD (cortical infarctions or multiple lacunes in strategic areas) were not included, as were subjects with other MRI abnormalities possibly interfering with the diagnosis. Additionally, the subjects did not meet the criteria for Alzheimer’s disease (AD) (McKhann et al., 1984). As a result of this selection procedure, all subjects had a severe global WMH score and no other suspected cause for the dementia syndrome. The controls were individuals from the same database who underwent the screening because of subjective cognitive complaints but who had no cognitive abnormalities on testing. Subjects were individually matched (one control for each VaD patient) for age and gender to the patients. The study groups were identical to the dataset used in a previous study on quantitative EEG (van Straaten et al., 2012). Table 1 shows demographic characteristics of the two groups.
EEG recordings At the first visit resting state EEG of 20 minutes was recorded in all subjects with OSG digital equipment (Brainlab®; OSG b.v. Belgium). Twenty-one electrodes were placed according to the 10–20 international system (Fp2, Fp1, F8, F7, F4, F3, A2, A1 T4, T3, C4, C3, T6, T5, P4, P3, O2, O1, Fz, Pz, and Cz). Electrode impedance was below 5 kΩ. Filter settings were: time constant 1 s, and low pass filter 70 Hz. Sample frequency was 500 Hz and analog–digital precision was 16 bit. EEGs were recorded (all electrodes separately and
dPLI in WMH – van Straaten et al. using an average reference). Subjects were seated in a slightly reclined chair in a soundattenuated, normally lit room and were kept awake as much as possible during the recording. For each subject, four epochs (average reference montage) of 10 seconds (5000 samples) each were visually selected on the basis of signal quality (artefact-free) and patient status (eyes closed, awake), then converted to ASCII format. The first 4096 samples of each epoch were used for further analysis (BrainWave analysis software, version 0.9.72, Stam, 2012; available at: http:/home.kpn.nl/stam7883/brainwave.html). EEGs were digitally rereferenced offline to the average A1/A2 reference derivation. Subsequently, a Hilbert transform was used to compare signals.
Functional connectivity analysis For each EEG epoch, we calculated the undirected functional connectivity with the Phase Lag Index (PLI). With this method, the distribution of differences of the instantaneous phases, determined from a Hilbert transform, between every channel pair is computed, resulting in a PLI value between 0 and 1 (0 ≤ PLI ≤ 1) for each channel pair:
PLI = sign[∆φ (tk )]
(1)
where ∆φ (tk ) , k = 1 . . . N is a time series of the phase differences.
Higher PLI values mean higher consistency of coupling, a value of 0 indicates no coupling or coupling with a phase difference of zero (which cannot reliably be distinguished from volume conduction effects) (Stam et al., 2007). A PLI value > 0 indicates a true interaction without noise from volume conduction and without bias from an active reference in the case of EEG. PLI values were calculated in five frequency bands (delta 0.5–4 Hz, theta 4–8 Hz, alpha1 8–10 Hz, alpha2 10-13 Hz, and beta 13–30 Hz) in an average reference derivation (using all electrodes except FP2/1 and A2/1). We left gamma band (> 30 Hz) activity out of the analyses because this fast activity cannot reliably be distinguished from muscle artefact (Whitham et al., 2007). PLI per subject was calculated by taking the mean of the four epochs. This was done for all EEG channels separately, as well as for the mean of all channels. For regional differences we clustered the channels to render six regions for further analysis, as used in a previous study: frontocentral right (Fp2, F4, C4), frontocentral left (Fp1, F3, C3), temporal right (F8, T4, T6), temporal left (F7, T3, T5), posterior right (P4, O2), and posterior left (P3, O1).
Measurement of direction and delay
dPLI in WMH – van Straaten et al.
In addition to the PLI, we computed for all the EEG epochs the directed Phase Lag Index (dPLI). This is a recently introduced connectivity measure that not only establishes the distribution of the instantaneous phase differences between channels (in the same manner as with the PLI), but also indicates what the direction of the differences is (Stam and van Straaten, 2012). To put differently: of every EEG electrode channel pair, in every time point, it is established which channel is leading and which channel is lagging in phase and how much by:
dPLIxy =
1 N
N
∑ H (φ ) t
(2)
t =1
where H is the Heaviside step function and
φt , t = 1 .... N a time series of phase differences
of time series x and y. Directed PLI ranges from 0 – 1 (0 ≤ dPLI≤ 1). For each channel pair xy, differences of the instantaneous phases, again determined from a Hilbert transform, in each time point are computed. The corresponding dPLI value is based on the direction of this difference, and a mean dPLI was computed for all channels separately by taking the mean of the four epochs. When no consisting phase leading or lagging is measured, the two channels are each leading in an equal amount. This will give a dPLI value of 0.5 by convention. When one channel is constantly leading compared to the other, this will result in a dPLI value for the leading channel of 1 and a dPLI value of almost 0 for the channel that is lagging. For illustration, Figure 1 shows the output of a dPLI computation of one epoch of a control subject. The dPLI cannot be averaged over all EEG channels, since this will result in a value of 0.5: the leading channels average out the lagging channels. For the computation and assessment of the differences between groups, we clustered the channels into the same six regions as described for the undirected connectivity: bilaterally frontocentral, temporal, and posterior.
Relationship with cognitive impairment In the patient group, connectivity (undirected and directed) was related to neuropsychological tests to assess the association between functional connectivity and cognitive performance. As a global cognitive measure we used the Mini Mental State Examination (MMSE) and for the assessment of memory function we used the Rey auditory verbal learning test (RAVLT, sum of five learning trials for immediate recall and a delayed recall trial after 20 minutes) (Folstein et al., 1975;Rey, 1958). Trailmaking test B (TMT B) and verbal fluency (animals) test were used as measures for executive functioning
dPLI in WMH – van Straaten et al.
(Reitan and Wolfson, 2004;Spreen and Straus, 1998). Missing values due to the subject’s cognitive incapability to perform or complete the test were set at the lowest possible score for that test.
Statistical analysis The distribution of functional connectivity values was checked with the KolmogorovSmirnov test. Small deviations from a normal distribution were accepted. Paired t-tests were used for comparison of mean PLI. No mean dPLI was tested since it follows from the method that the mean dPLI over all channels is 0.5. To test for group and regional differences in dPLI, a repeated-measures analysis of variance (ANOVA) with GreenhouseGeisser correction for sphericity was performed, with group as the between subject variable and the region (frontocentral, temporal, and posterior; right and left) as the within subject variable. In frequency bands with a significant difference between groups or with a significant interaction between group and area in the ANOVA, a post-hoc permutation test was performed to display the brain areas with significant differences between the groups: Directed PLI differences between the groups were compared to differences of two groups with randomly assigned subjects (repeated 100 times). In the patient group the EEG measures per frequency band were related to the cognitive measures using linear regression analysis with PLI and dPLI (of region frontocentral right) as independent variables and MMSE, RAVLT, TMT B and verbal fluency as the dependent variables.
Results
General characteristics and functional connectivity In total, seventeen WMH-VaD patients were identified meeting inclusion criteria. Three individuals in the patient group and two in the control group had CT available instead of MRI. PLI values of the WMH-VaD patients did not differ significantly from those of the controls in any band (Table 2). Also, no interaction between the regional PLI and group was found (F ≤ 0.80, p > 0.05).
Direction and delay Directed PLI differed significantly between groups. Group differences were found with ANOVA in the beta band (F 7.71, p 0.5 (F3 leading over O1) and all lines under the x-axis (negative y-value) correspond to a dPLI value < 0.5 (O1 leading over F3). More lines are located above the xaxis, and this is confirmed by the mean dPLI value of the whole epoch: 0.77. C. Colourcoded dPLI values for all channels for the first 50% of the epoch. X-axis: time. Y-axis: electrodes (numbers correspond with the order in which the electrodes are mentioned in paragraph 2.2). Blue colour refers to a dPLI value < 0.5, red colour indicates dPLI > 0.5. The posterior regions (electrodes 13 – 18, and 21 are relatively more blue (lagging) compared to the anterior regions (electrodes 1 – 12, 19, and 20). D. dPLI anatomically projected (nose up, left = left hemisphere), red is high dPLI, blue is low dPLI. Lines are connections between brain areas, based on the Phase Lag Index (PLI). A front-to-back (from red to blue) distribution of dPLI can be seen. The colour bar applies to both Figure 1C and D.
Figure 2. Topographical representation (nose up, left = left hemisphere) of average directed Phase Lag Index (dPLI) for each group. Red indicates relatively high value, blue indicates relatively low value. Connections between brain regions based on the Phase Lag Index are indicated with lines. Stars indicate brain area with difference between patients and controls (post-hoc permutation analysis, p 0 and dPLI between 0 and 0.5 in the channel that is lagging and dPLI between 0.5 and 1 in the channel that is leading. B: presumed situation in WMH-VaD with phase differences between channels ≠ 0, and with alternatingly leading and lagging of channels. This corresponds to a PLI > 0, and dPLI close to 0.5.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
dPLI in WMH – van Straaten et al.
Table 1. Characteristics of white matter intensities based vascular dementia patients and control subjects. Patients (n=17)
Control subjects (n=17)
Age (yrs )
74.6 ± 8
74.4 ± 8
Sex ratio (m/f)
9/8
9/8
Mean WMH score (SD)
3 ± 0*
1 ± 1*
Mean MMSE score (SD)
22 ± 5*
28.5 ± 1*
Mean MTA score (SD)
1.7 (0.6)
1.0 (0.8)
MMSE: mini mental state examination, MTA: medial temporal lobe atrophy, WMH: white matter hyperintensities. *: p < 0.05 for difference between groups
dPLI in WMH – van Straaten et al.
Table 2. Functional connectivity values for the three subject groups.
Mean PLI control subjects (SD)
Mean PLI VaD patients (SD)
Delta
Theta
Alpha 1
0.15 (0.02)
0.17 (0.07)
0.25 (0.08)
0.15 (0.03)
0.17 (0.04)
0.22 (0.07)
N = 17 in each group. PLI: Phase Lag Index.
Alpha 2
Beta
0.15
0.08
(0.03)
(0.01)
0.16
0.07
(0.03)
(0.03)
dPLI in WMH – van Straaten et al.
Table 3. Analysis of variance (repeated measures) for regional directed Phase Lag Index values in each frequency band. Within subjects Area
Area x Group
Between subjects Group
Delta band
F [3.11]=4.38**
F [3.11]=2.22
F [1]=4.59*
Theta band
F [1.66]=2.13
F [1.66]=2.34
F [1]=4.16*
Alpha 1 band
F [2.00]=11.90**
F [2.00]=3.28*
F [1]=5.36*
Alpha 2 band
F [2.45]=1.52
F [2.45]=4.09*
F [1]=1.68
Beta band
F [2.19]=2.27
F [2.19]=4.82**
F [1]=7.71**
* p < 0.05 ** p < 0.01 Degrees of freedom are printed between square brackets.
