Relations between structural and EEG-based graph metrics in healthy controls and schizophrenia patients

MRI and EEG networks in schizophrenia Gomez Pilar et al. 1 RELATIONS BETWEEN STRUCTURAL AND EEG-BASED GRAPH METRICS IN HEALTHY CONTROLS AND SCHIZOPHRENIA PATIENTS Running title: MRI and EEG networks in schizophrenia Javier Gomez-Pilar a, Rodrigo de Luis-García b, Alba Lubeiro c, Henar de la Red d , Jesús Poza a,d,e,f, Pablo Núñez a, Roberto Hornero a,e,f , Vicente Molina c,d,e * a Biomedical Engineering Group, University of Valladolid, Paseo de Belén, 15, 47011 Valladolid, Spain b Imaging Processing Laboratory, University of Valladolid, Paseo de Belén, 15, 47011 Valladolid, Spain. c Psychiatry Department, School of Medicine, University of Valladolid, Av. Ramón y Cajal, 7, 47005 Valladolid, Spain. d Psychiatry Service, Clinical Hospital of Valladolid, Ramón y Cajal, 3, 47003 Valladolid, Spain. e Neurosciences Institute of Castilla y León (INCYL), Pintor Fernando Gallego, 1, 37007 University of Salamanca, Spain. f IMUVA, Mathematics Research Institute, University of Valladolid, Valladolid, Spain. MRI and EEG networks in schizophrenia Gomez Pilar et al. 2 SUMMARY Objective: To assess using graph-theory properties of both structural and functional networks in schizophrenia patients, as well as the possible prediction of the latter based on the former. Abnormal structural and functional network parameters have been found in schizophrenia, but the dependence of functional network properties on structural alterations has not been described yet in this syndrome. Experimental design: We applied averaged path-length (PL), clustering coefficient (CLC) and density (D) measurements to structural data derived from diffusion magnetic resonance and functional data derived from electroencephalography in 39 schizophrenia patients and 79 controls. Functional data were collected for the global and theta frequency bands with subjects performing an odd-ball task, both prior to stimulus delivery and at the corresponding processing window. Connectivity matrices were constructed respectively from (i) tractography and registered cortical segmentations (structural) and (ii) phase-locking values (functional). Principal observations: In both groups, we observed a significant EEG task-related modulation (change between pre-stimulus and response windows) in the global and theta bands. Patients showed larger structural PL and pre-stimulus density in the global and theta bands, and lower PL task-related modulation in the theta band. Structural network values predicted pre-stimulus global band values in controls and global band task-related modulation in patients. Abnormal functional values found in patients (prestimulus density in the global and theta bands and task-related modulation in the theta band) were not predicted by structural data in this group. Structural and functional network abnormalities respectively predicted cognitive performance and positive symptoms in patients. Conclusions: Taken together, the alterations in the structural and functional theta networks in the patients and the lack of significant relations between these alterations, suggest that these types of network abnormalities exist in different groups of schizophrenia patients. Key words: schizophrenia, dysconnectivity, network, graph-theory, electroencephalography, diffusion magnetic resonance. MRI and EEG networks in schizophrenia Gomez Pilar et al. 3 1. INTRODUCTION Mental functions depend on global dynamics of cerebral networks (Dehaene and Changeux, 2011; Varela, et al., 2001), whose functional and structural characteristics can be assessed in vivo using methods derived from graph-theory (Bullmore and Sporns, 2009). In this context, underpinnings of syndromes like schizophrenia likely involve distributed networks rather than regional alterations, as supported by studies using functional magnetic resonance imaging (fMRI) that revealed network alterations in the resting state (Lo, et al., 2015; Yu, et al., 2011) and during task performance (Ma, et al., 2012; Shim, et al., 2014) in this syndrome. However, considering the rapid and transient change of functional integration of diverse cerebral regions in cognition in humans (Varela, et al., 2001) and animals (Bressler, et al., 1993), assessing fast change of cerebral networks in schizophrenia holds a great interest. Techniques with high temporal resolution are useful to this purpose: change of network properties using EEG during a cognitive task was significantly decreased in schizophrenia patients (Gomez-Pilar, et al., 2017a). Using relative power analyses, we also reported lower EEG task-related change in theta but not in faster bands during an odd-ball task in schizophrenia (Bachiller, et al., 2014). As mentioned, methods derived from graph-theory are useful to assess the properties of cerebral networks, which can be summarized in parameters such as clustering coefficient (CLC) and characteristic path length (PL). In a binary network, local CLC is the ratio between the number of triangles in which a given node participates and the maximum possible number of triangles including that node. When CLC is averaged across the nodes of a network, it quantifies network segregation and local efficiency of information transfer. In turn, PL is the average of shortest distances for all possible pairs of nodes; it is likely related to information integration across areas. These network parameters provide complementary information about the properties of the whole brain network. Therefore, the use of these parameters instead of their corresponding nodal versions, allows to characterize the global and predominant changes of the network. A recent meta-analysis of functional graph-analytical studies in schizophrenia revealed significant decreases in measures of local organization (CLC) with preservation in short communication paths (PL) (Kambeitz, et al., 2016). Abnormalities in structural connectivity are also prevalent in schizophrenia (Ellison- Wright and Bullmore, 2009). These abnormalities are likely reflected in structural network properties, since longer structural PL values were found in schizophrenia at frontal and temporal regions using dMRI (van den Heuvel, et al., 2010) and may be associated to genetic liability to this disorder (Bohlken, et al., 2016). Thus, the possibility exists that functional network alterations might be secondary to structural abnormalities in schizophrenia. Indeed, in this syndrome, a relationship has been MRI and EEG networks in schizophrenia Gomez Pilar et al. 4 reported between a reduction in “rich-club density” (i.e., connections among highdegree hub nodes) and global efficiency of functional connectivity in the resting state using fMRI (van den Heuvel, et al., 2013). Similarly, connectivity deficits in rich-club hubs have been described in young offspring of schizophrenia patients associated to disruption of the functional connectome (Collin, et al., 2017). However, functional connectivity alterations in schizophrenia are not necessarily determined by structural connectivity, since functional connections in the resting state can be found between regions without direct anatomical connections (Honey, et al., 2009). The application of graph-theory parameters to functional and structural measurements can yield complementary information and help uncovering hidden relationships (Sui, et al., 2012). Using diffusion magnetic resonance imaging (dMRI), graph-theory parameters may inform about structural connectivity differences between anatomical structures, revealing highly connected hubs (Honey, et al., 2010). Graph-theory parameters applied to functional analysis may reveal baseline network characteristics and its dynamic modulation during cognition of signals such as synchrony of the bold-oxygen level dependent signal between regions, or magneto-electrical signals between sensors. Considering the millisecond-scale of modulation of cortical activity during cognition (Bressler, et al., 1993; Dehaene and Changeux, 2011), the combination of network analyses with temporal resolution of electroencephalographic (EEG) recordings can be useful to assess this task-related modulation. Indeed, using EEG in healthy subjects, we reported a significant task-related modulation of network parameters from pre-stimulus (from -300 to 0 ms prior to stimulus onset) to response (from 150 to 450 ms poststimulus) windows (Martin-Santiago, et al., 2016) during an odd-ball task. To our knowledge, no previous study has assessed the relationship between structural and EEG networks in schizophrenia. Such investigation may help identifying the substrate of the cortical dysfunction in schizophrenia. Therefore, the present study was aimed at characterizing the properties of structural and EEG-based functional networks in schizophrenia and assessing the relationships between properties of those networks in this syndrome, particularly between structural connectivity and EEG modulation. 2. SUBJECTS AND METHODS 2.1 Subjects Thirty-nine schizophrenia (19 stable chronic and 20 first-episode, FE) patients and 78 healthy controls with normal hearing were included. Demographic, clinical, cognitive and EEG data were collected for each participant (Table 1). In addition, dMRI data were also available in 33 patients (16 FE) and 27 controls (Table 1). One of the psychiatrists in the group (VM) diagnosed the patients according to the Diagnostic and Statistical Manual of Mental Disorders, 5th edition. Chronic patients received atypical antipsychotics, 30 of them in monotherapy (12 received antidepressants and 7 MRI and EEG networks in schizophrenia Gomez Pilar et al. 5 benzodiazepines). FE patients were receiving stable doses of antipsychotics for less than 15 days, with a wash-out period of 24 hours prior to EEG acquisition. Symptoms were scored using the Positive and Negative Syndrome Scale (PANSS) (Kay, et al., 1987). Exclusion criteria were: (i) any neurological illness; (ii) history of cranial trauma with loss of consciousness longer than one minute; (iii) past or present substance abuse, except nicotine or caffeine; (iv) total intelligence quotient (IQ) smaller than 70; and (iv) for patients, any other psychiatric process, and (v) for controls, any current psychiatric or neurological diagnosis or treatment. The population here included overlaps in part with that of previous reports of our group in schizophrenia on functional networks based on evoked response (Gomez-Pilar, et al., 2017a), graph complexity (Gomez-Pilar, et al., 2017b) and structural connectivity of specific tracts of the prefrontal region (Molina, et al., 2017) We obtained written informed consent from all participants after full printed information. The ethical committees of the University Hospital of Valladolid approved the study. 2.2. Cognitive assessment Cognitive data from patients and controls were collected using: the Wechsler Adult Intelligence Scale, WAIS-III (IQ); the Wisconsin Card Sorting Test (WCST; completed categories and percentage of perseverative errors); and the Spanish version of the Brief Assessment in Cognition in Schizophrenia Scale (BACS) (Segarra, et al., 2011). 2.3. MRI acquisition and processing Acquisitions were carried out using a Philips Achieva 3 Tesla MRI unit (Philips Healthcare, Best, The Netherlands) at the MRI facility at Valladolid University, including anatomical T1-weighted and diffusion-weighted images. For the T1-weighted images, acquisition parameters were: turbo field echo (TFE) sequence, 256 x 256 matrix size, 1 x 1 x 1 mm3 of spatial resolution and 160 slices covering the whole brain. About the diffusion weighted images, the acquisition protocol parameters were: 61 gradient directions and one baseline volume, b-value = 1000 s/mm2, 2 × 2 × 2 mm3 of voxel size, 128 × 128 matrix and 66 slices covering the entire brain. Total acquisition time was 18 minutes. The processing pipeline of the acquired MRI volumes is designed to obtain structural connectivity matrices by using both the anatomical (T1-weighted) and diffusion images (Fig. 1). First, non-brain structures were removed from the T1 images, using BET, the brain extraction tool from the FSL software suite (http://fsl.fmrib.ox.ac.uk) (Smith, 2002). After that, the segmentation of 84 cortical structures was performed employing Freesurfer (https://surfer.nmr.mgh.harvard.edu) (Desikan, et al., 2006; Fischl, et al., 2004). From the same T1 images, gray matter, white matter and cerebrospinal fluid (CSF) were also segmented, and subcortical gray matter structures were obtained using FAST MRI and EEG networks in schizophrenia Gomez Pilar et al. 6 and FIRST utilities from FSL, respectively (Patenaude, et al., 2011; Zhang, et al., 2001), and combined into a volume called 5tt (5-tissue-type) image. In parallel, the brain was extracted from the diffusion weighted images (DWI) using DWI2MASK tool from MRtrix (www.mrtrix.org) (Dhollaner and Connelly, 2016). Also employing MRtrix, orientation distribution functions were estimated from the diffusion data using spherical deconvolution(Tournier, et al., 2007), which were later employed to generate anatomically-constrained tractography using both the diffusion data and the 5tt image (after registration). Two million streamlines were generated for each subject. In order to characterize diffusion at each voxel, diffusion tensors were estimated using a least squares method (Salvador, et al., 2005), and scalar Fractional Anisotropy (FA) volumes were computed from the diffusion tensors. FA quantifies the amount of anisotropy in the diffusion tensor, that is, how much it deviates from a totally isotropic diffusion. FA is usually interpreted as a descriptor of white matter integrity, and decreases in FA have been related to alterations in the white matter due to several factors (demyelination and axonal destruction, among others). Finally, connectivity matrices were constructed from the tractography results and the (registered) cortical segmentations. When a streamline between two cortical segmentations was found, the averaged FA was computed. Thus, 84x84 connectivity matrices were obtained using FA as connectome metrics (Fig. 2). A threshold was not applied to the obtained matrices; however, some matrix coefficients were equal to zero when a streamline was not found. Similar connectomics analyses have been reported in schizophrenia (Di Biase, et al., 2017) and other neurocognitive conditions (Jones, et al., 2015) 2.4. EEG recordings and processing 2.4.1 EEG acquisition and preprocessing EEG recordings were obtained following MRI scans, after a resting period of 30 minutes. Participants performed a 13 minutes three-tone P300 oddball task (for a detailed description see (Gomez-Pilar, et al., 2017a). Electrode impedance was always kept under 5 kΩ. Ground was placed at Fpz electrode and each channel was referenced over Cz electrode and re-referenced to the average activity of all active sensors (Bledowski, et al., 2004; Gomez-Pilar, et al., 2017b), yielding a total of 29 channels. The P300 task has several advantages for assessing functional network modulation in schizophrenia. In addition to its widespread previous use in the field: (i) it is easy to perform, thus decreasing bias related to lack of subject’s cooperation; (ii) its performance activates a large cerebral network (Linden, et al., 1999; Bledowski, et al., 2004); and (iii) differences in EEG global activation patterns have been reported in schizophrenia using this paradigm (Gomez-Pilar, et al., 2017a). MRI and EEG networks in schizophrenia Gomez Pilar et al. 7 Signals were band pass filtered between 1 and 70 Hz. In addition, a zero-phase 50 Hz notch filter was used to remove the power line artifact. A three-step artifact rejection algorithm was applied to minimize electroculographic and electromiographic contamination (Bachiller, et al., 2015a): (i) an Independent Component Analysis (ICA) was carried out to discard noisy ICA components; (ii) after ICA reconstruction, EEG signals were divided into trials of 1 second length (ranging from 300 ms before to 700 ms after stimulus onset); and (iii) an automatic method was applied to reject trials whose amplitude exceeded an adaptive statistical-based threshold, which consists of two stages. First, the mean and standard deviation of each channel was computed. Then, trials that exceeded mean ± 4 × standard deviation in at least two channels were discarded (Nunez, et al., 2017). After this adaptive artifact rejection, 91.21 ± 11.28 trials for controls and 86.33 ± 14.13 for patients were left for further analyses. In order to describe the event-related potential (ERP) waveforms, Figure 1S has been include in the Supplementary material. ERPs in the midline electrodes are shown in Figure S1.A. Figure S1.B shows the channel grand average waveforms. Finally, Figure S1.C depicts scalps maps with the P300 peak amplitude for both groups. 2.4.2 EEG brain graphs The functional brain network was characterized using EEG graphs. Electrodes were used to represent network nodes, whereas network edges were set by computing the neural coupling between pairs of electrodes. Specifically, neural coupling was established using the phase locking value (PLV) across successive trials (Lachaux, et al., 1999). PLV in sensible to small oscillations of the EEG (Spencer, et al., 2003) and takes into account nonlinearities (van Diessen, et al., 2015), which is an intrinsic feature of EEG recordings. PLV can be computed using different approaches. In this study, the continuous wavelet transform (CWT) was used to extract the phase information from each trial (Bob, et al., 2008). Edge effects in CWT were considered by computing the cone of influence (COI) for pre-stimulus and response time windows. Only wavelet coefficients inside the respective COI were considered for the analyses to avoid edge effects. We refer to our previous studies (Gomez-Pilar, et al., 2017b) for detailed explanations about how wavelet coefficient were computed, the wavelet parameters were configured and the COIs were applied to the CWT decomposition in order to minimize edge effects. After using CWT approach to perform filtering and phase extraction in one operation (Bob et al., 2008), the PLV between two signals x(t) and y(t) can be obtained evaluating the variability of the phase difference across successive trials: PLVxy(k,s)=1 Nt |∑eφxy(k,s,n) N n=1 |, (1) where Nt is the number of trials, φxy is the instantaneous phase difference between x and y signals, k is the time interval, and s the scaling factor of the mother wavelet (see (Bachiller et al., 2015) for details). MRI and EEG networks in schizophrenia Gomez Pilar et al. 8 Thus, functional connectivity matrices based on PLV ranged between 0 and 1, where a value of 1 is obtained with completely synchronized signals and a value of 0 implies an absence of synchronization. Following the same methodology as in the structural data, functional connectivity matrices were not thresholded. 2.4.3 Segmentation of the EEG response In order to assess the task-related modulation of the graph parameters along the oddball task, two time windows were considered for comparison. On the one hand, the prestimulus window (i.e. a period of expectation before the stimulus onset) ranges from - 300 ms to the stimulus onset. On the other hand, the response window was selected to capture the P3b response. In order to take into account the inter-individual variability of the P3b response, the response window was adaptively set for each participant. Firstly, the event-related wave was computed for each subject by the synchronized averaging of all the trials corresponding to attended target tones. Secondly, a low-pass finite impulse response filter with cut-off frequency of 8 Hz was applied to the evoked wave in order to obtain only the components related to delta and theta frequency bands. It is noteworthy that this filter was only applied to estimate the time window related to the EEG response. Thirdly, the maximum amplitude of the low-pass filtered evoked wave in the Pz channel was located into a window ranging from 250 to 550 ms from the stimulus onset (Bachiller, et al., 2015b). The corresponding sample to the maximum amplitude was used as a central time sample of the response window. Finally, the response window was set on ±150 ms around the central time sample. 2.5. Graph-theory parameters From both the structural and functional connectivity matrices, we calculated three graph-theory parameters to characterize global properties of the brain network: (i) connectivity strength (i.e. mean network degree) by means of network density (D), also named network strength (ii) network segregation using CLC, and (iii) network integration by means of PL (Rubinov and Sporns, 2010). For the sake of comparability and to obtain results independent of network size and network strength,, CLC and PL were computed over an ensemble of 50 surrogate random networks, which were used to normalize CLC and PL values obtained from the original networks (Stam, et al., 2009). Therefore, normalized CLC and PL can be defined: CLC = C Crandom, (2) PL =L Lrandom, (3) where C and L can be defined as follows: MRI and EEG networks in schizophrenia Gomez Pilar et al. 9 C = 1 N∑ ∑∑wijwilwjli≠l j≠l i≠j ∑ ∑ wijwili≠l j≠l i≠j N i=1 , (4) L = N(N−1) ∑ ∑ 1 Lij N j≠i N i=1 , (5) In equation (4), wij can be referred to PLV between nodes i and j (for functional analyses) or the structural connectivity between two regions using the streamlines from MRI. N is the total number of nodes of the network (29 in EEG analyses, 84 in MRI). Finally, Lij is defined as the inverse of the edge weight (Stam et al., 2009). With regard to the EEG functional network, parameters were computed into two frequency ranges. They were selected based on their relevance for the task-related modulation of the EEG during P300 tasks shown in schizophrenia in previous studies: the global band (from 1 to 70 Hz) (Gomez-Pilar, et al., 2017a) and the theta band (from 4 to 8 Hz) (Bachiller, et al., 2014; Doege, et al., 2009). A diminished task-related modulation of theta activity during an oddball task was found in schizophrenia, but not in faster frequency bands (Bachiller et al., 2014). In addition, the assessment of the theta band showed abnormalities in the brain network reconfiguration in the secondary functional pathways in schizophrenia (Gomez-Pilar et al., 2017b). On the other hand, the global band could be also useful to assess the specificity of the theta band. Functional network parameters during pre-stimulus and its corresponding task-related modulation (i.e. difference between the response and the pre-stimulus windows) were used for statistics. Structural connectivity network parameters will be referred to as dMRI-PL, dMRI-D and dMRI-CC, and functional network parameters as EEG-PL, EEG-D and EEG-CLC. 2.6. Statistics We compared socio-demographic data (age, sex, education years and parental education) between patients and controls (t or χ2 tests when appropriate). Each subgroup of patients (i.e., those with only EEG and those with EEG plus dMRI data) was compared to the corresponding controls. 2.6.1 Comparisons of graph-theory parameters After testing normality and homoscedasticity of data distribution using Kolmogorov- Smirnov and Levene tests, we compared functional (EEG-based) and structural (dMRI- based) graph-theory parameters between patients and controls using Student´s t-tests. Within-group changes in functional network parameters were assessed using t-tests for related samples. After Bonferroni adjustment, p level was set to 0.05/15=0.003. MRI and EEG networks in schizophrenia Gomez Pilar et al. 16 would be not closely associated to anatomical connectivity deficits according to our results. Such relative independence of structural and functional connectivity alterations surprised us, but could be explained by data showing that functional connectivity exists between regions without direct anatomical connection (Adachi, et al., 2012; Honey, et al., 2009). Thus, deficits of functional integration (in the theta band) would not require altered structural substrates. Coordination of activity between distant regions may be established indirectly, since functional connectivity is high among regions with common efferences to third regions, which may convey information to higher regions and may also receive similar afferences (Adachi, et al., 2012). Therefore, the alteration in relevant cortical hubs reported in schizophrenia (van den Heuvel, et al., 2013) may hamper the synchronization of regions not directly linked via with matter tracts. Although other data using anatomical and functional MRI show a substantial correspondence between the corresponding networks (Hagmann, et al., 2008), this relation had not been assessed yet with EEG data. Considering all this, we must underline that structural deficits were found in our patients (larger dMRI-PL and lower factor scores) and were predictive of positive symptoms. This suggests the coexistence of alterations in both structural and functional networks (in the theta band) within schizophrenia, but not necessarily in the same cases. In other words, either both unrelated functional and structural networks alterations are found in schizophrenia or they are characteristics of different schizophrenia subgroups. The latter possibility seems favored by recent reports supporting that structural connectivity values can segregate biologically valid clusters within schizophrenia (Lubeiro, et al., 2016; Sun, et al., 2015; Wheeler, et al., 2015). Using EEG, both no difference (Jhung, et al., 2013; Rubinov, et al., 2009) and a decrease (Micheloyannis, et al., 2006) of CLC at rest were reported in schizophrenia, which may be coherent with that possibility. Remarkably, pre-stimulus EEG-D is higher in the patients. The density is the mean network degree (i.e. a measure of the network strength), implying a functional overconnectivity at rest in schizophrenia. This result is in agreement with the increased prefrontal functional connectivity reported in schizophrenia (Anticevic, et al., 2015). The different patterns of dMRI-D and EEG-D in patients, and the lack of a significant relation between them, support the independence of the alterations in both networks. Speculatively, the increased EEG-D might relate to the deficit in GABA function observed in schizophrenia (Gonzalez-Burgos, et al., 2011; Thakkar, et al., 2017), which could lead to hyper-synchronization. In our study, functional connectivity is based on PLV values; thus, larger pre-stimulus theta EEG-D values suggest and excess of synchronization in the patients in this band, which could have a ceiling effect on task-related synchronization and might hamper theta EEG-D modulation. Therefore, an inhibitory transmission deficit could justify both the increased baseline D values and the lower modulation in the theta band, given its large implication in P300 task performance (Bachiller, et al., 2014; Doege, et al., 2009). This possible dependence on inhibitory function might also justify the lack of a significant prediction of theta modulation by structural connectivity. MRI and EEG networks in schizophrenia Gomez Pilar et al. 17 The increase in theta task-related modulation values (i.e., larger functional density, CLC and PL in this band) predicted better cognition in the patients. There was only one cognitive factor, which is not surprising since the assessment instrument (BACS) included the dimensions where performance was previously found decreased in schizophrenia. That predictive relationship suggests that cognitive deficit is secondary to the decreased capacity of modulating the functional network in the theta band, perhaps indicating a lesser capacity to integrate the activity of different areas in a task. Our study is limited by the sample size of patients with both structural and functional network data available. A larger sample would be needed to test the hypothesis of distinct schizophrenia clusters based on structural connectivity. In addition, the assessment of nodal parameters could be of interest. However, connectivity analysis at the sensor level is very problematic due to effects of field spread (Schoffelen and Gross, 2009). Therefore, future studies should also be focused on increasing the number of EEG electrodes to provide more accurate results. Moreover, we cannot rule out an effect of treatment, although antipsychotic doses were unrelated to structural and functional graph parameters. It must be also noted that all EEG measures are influenced by volume conduction. In order to minimize this effect, a well-known strategy is based on the assumption that volume conduction affects the connectivity estimates in a similar way in two different experimental contrasts, such as pre-stimulus and response conditions (Bastos and Schoffelen, 2016). With regard to the use of dMRI-based connectivity, the accuracy of the cortical segmentation and the choice of the tractography method influence the obtained connectivity matrices. Although FA is the most usual dMRI descriptor for white matter integrity, it cannot identify the ultimate origin of connectivity alterations. We may conclude that task-related modulation deficit in the theta band in schizophrenia is independent from deviation from normal structural network properties. This, considered together with the different correlates of functional and structural connectivity alterations, might support different clusters within the schizophrenia syndrome. MRI and EEG networks in schizophrenia Gomez Pilar et al. 18 Conflicts of interest: None Acknowledgements This research project was supported in part by grants from Instituto de Salud Carlos III under project PI15/00299, “Gerencia Regional de Salud de Castilla y León” under projects GRS 1263/A/16 and GRS 1485/A/17, and “Ministerio de Economía y Competitividad” and FEDER under grants TEC2014-53196-R and TEC2013-44194-P; by ‘European Commission’ and FEDER under project 'Análisis y correlación entre el genoma completo y la actividad cerebral para la ayuda en el diagnóstico de la enfermedad de Alzheimer' ('Cooperation Programme Interreg V-A Spain-Portugal POCTEP 2014-2020'), and by ‘Consejería de Educación de la Junta de Castilla y León’ and FEDER under project VA037U16. J. Gomez-Pilar was in receipt of a grant from University of Valladolid and A. Lubeiro was in receipt of a grant from Consejería de Educación de la Junta de Castilla y León. MRI and EEG networks in schizophrenia Gomez Pilar et al. 19 5. REFERENCES Adachi, Y., Osada, T., Sporns, O., Watanabe, T., Matsui, T., Miyamoto, K., Miyashita, Y. (2012) Functional connectivity between anatomically unconnected areas is shaped by collective network-level effects in the macaque cortex. Cereb Cortex, 22:1586-92. Anticevic, A., Hu, X., Xiao, Y., Hu, J., Li, F., Bi, F., Cole, M.W., Savic, A., Yang, G.J., Repovs, G., Murray, J.D., Wang, X.J., Huang, X., Lui, S., Krystal, J.H., Gong, Q. (2015) Early-course unmedicated schizophrenia patients exhibit elevated prefrontal connectivity associated with longitudinal change. J Neurosci, 35:267-86. Bachiller, A., Diez, A., Suazo, V., Dominguez, C., Ayuso, M., Hornero, R., Poza, J., Molina, V. (2014) Decreased spectral entropy modulation in patients with schizophrenia during a P300 task. Eur Arch Psychiatry Clin Neurosci, 264:533-43. Bachiller, A., Poza, J., Gomez, C., Molina, V., Suazo, V., Hornero, R. (2015a) A comparative study of event-related coupling patterns during an auditory oddball task in schizophrenia. J Neural Eng, 12:016007. Bachiller, A., Romero, S., Molina, V., Alonso, J.F., Mananas, M.A., Poza, J., Hornero, R. (2015b) Auditory P3a and P3b neural generators in schizophrenia: An adaptive sLORETA P300 localization approach. Schizophr Res, 169:318-25. Bastos, A.M., Schoffelen, J.M. (2016) A Tutorial Review of Functional Connectivity Analysis Methods and Their Interpretational Pitfalls. Front Syst Neurosci, 9:175. Bledowski, C., Prvulovic, D., Hoechstetter, K., Scherg, M., Wibral, M., Goebel, R., Linden, D.E. (2004) Localizing P300 generators in visual target and distractor processing: a combined event-related potential and functional magnetic resonance imaging study. J Neurosci, 24:9353-60. Bob, P., Palus, M., Susta, M., Glaslova, K. (2008) EEG phase synchronization in patients with paranoid schizophrenia. Neurosci Lett, 447:73-7. Bohlken, M.M., Brouwer, R.M., Mandl, R.C., Van den Heuvel, M.P., Hedman, A.M., De Hert, M., Cahn, W., Kahn, R.S., Hulshoff Pol, H.E. (2016) Structural Brain Connectivity as a Genetic Marker for Schizophrenia. JAMA Psychiatry, 73:11-9. Bressler, S.L., Coppola, R., Nakamura, R. (1993) Episodic multiregional cortical coherence at multiple frequencies during visual task performance. Nature, 366:153-6. Brunner, C., Billinger, M., Seeber, M., Mullen, T.R., Makeig, S. (2016) Volume Conduction Influences Scalp-Based Connectivity Estimates. Front Comput Neurosci, 10:121. Bullmore, E., Sporns, O. (2009) Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci, 10:186-98. Collin, G., Scholtens, L.H., Kahn, R.S., Hillegers, M.H.J., van den Heuvel, M.P. (2017) Affected Anatomical Rich Club and Structural-Functional Coupling in Young Offspring of Schizophrenia and Bipolar Disorder Patients. Biol Psychiatry. Dehaene, S., Changeux, J.P. (2011) Experimental and theoretical approaches to conscious processing. Neuron, 70:200-27. Desikan, R.S., Segonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker, D., Buckner, R.L., Dale, A.M., Maguire, R.P., Hyman, B.T., Albert, M.S., Killiany, R.J. (2006) An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage, 31:968-80. Dhollaner, T., Connelly, A. (2016) Unsupervised 3-tissue response function estimation from single-shell or multi-shell diffusion mr data without a co-registered t1 image. ISMRM Workshop on Breaking the Barriers of Diffusion MRI. Di Biase, M.A., Cropley, V.L., Baune, B.T., Olver, J., Amminger, G.P., Phassouliotis, C., Bousman, C., McGorry, P.D., Everall, I., Pantelis, C., Zalesky, A. (2017) White matter connectivity disruptions in early and chronic schizophrenia. Psychol Med:1-14. MRI and EEG networks in schizophrenia Gomez Pilar et al. 20 Doege, K., Bates, A.T., White, T.P., Das, D., Boks, M.P., Liddle, P.F. (2009) Reduced event-related low frequency EEG activity in schizophrenia during an auditory oddball task. Psychophysiology, 46:566-77. Ellison-Wright, I., Bullmore, E. (2009) Meta-analysis of diffusion tensor imaging studies in schizophrenia. Schizophr Res, 108:3-10. Fischl, B., van der Kouwe, A., Destrieux, C., Halgren, E., Segonne, F., Salat, D.H., Busa, E., Seidman, L.J., Goldstein, J., Kennedy, D., Caviness, V., Makris, N., Rosen, B., Dale, A.M. (2004) Automatically parcellating the human cerebral cortex. Cereb Cortex, 14:11-22. Gomez-Pilar, J., Lubeiro, A., Poza, J., Hornero, R., Ayuso, M., Valcarcel, C., Haidar, K., Blanco, J.A., Molina, V. (2017a) Functional EEG network analysis in schizophrenia: Evidence of larger segregation and deficit of modulation. Prog Neuropsychopharmacol Biol Psychiatry. Gomez-Pilar, J., Poza, J., Bachiller, A., Gomez, C., Nuñez, P., Lubeiro, A., Molina, V., Hornero, R. (2017b) Quantification of Graph Complexity Based on the Edge Weight Distribution Balance: Application to Brain Networks. Int J Neural Syst, in press. Gonzalez-Burgos, G., Fish, K.N., Lewis, D.A. (2011) GABA neuron alterations, cortical circuit dysfunction and cognitive deficits in schizophrenia. Neural Plast, 2011:723184. Griffa, A., Baumann, P.S., Ferrari, C., Do, K.Q., Conus, P., Thiran, J.P., Hagmann, P. (2015) Characterizing the connectome in schizophrenia with diffusion spectrum imaging. Hum Brain Mapp, 36:354-66. Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C.J., Wedeen, V.J., Sporns, O. (2008) Mapping the structural core of human cerebral cortex. PLoS Biol, 6:e159. Honey, C.J., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J.P., Meuli, R., Hagmann, P. (2009) Predicting human resting-state functional connectivity from structural connectivity. Proc Natl Acad Sci U S A, 106:2035-40. Honey, C.J., Thivierge, J.P., Sporns, O. (2010) Can structure predict function in the human brain? Neuroimage, 52:766-76. Jhung, K., Cho, S.H., Jang, J.H., Park, J.Y., Shin, D., Kim, K.R., Lee, E., Cho, K.H., An, S.K. (2013) Small-world networks in individuals at ultra-high risk for psychosis and first-episode schizophrenia during a working memory task. Neurosci Lett, 535:35-9. Jones, J.T., DiFrancesco, M., Zaal, A.I., Klein-Gitelman, M.S., Gitelman, D., Ying, J., Brunner, H.I. (2015) Childhood-onset lupus with clinical neurocognitive dysfunction shows lower streamline density and pairwise connectivity on diffusion tensor imaging. Lupus, 24:1081-6. Kambeitz, J., Kambeitz-Ilankovic, L., Cabral, C., Dwyer, D.B., Calhoun, V.D., van den Heuvel, M.P., Falkai, P., Koutsouleris, N., Malchow, B. (2016) Aberrant Functional Whole-Brain Network Architecture in Patients With Schizophrenia: A Meta-analysis. Schizophr Bull, 42 Suppl 1:S13-21. Kay, S.R., Fiszbein, A., Opler, L.A. (1987) The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull, 13:261-76. Lachaux, J.P., Rodriguez, E., Martinerie, J., Varela, F.J. (1999) Measuring phase synchrony in brain signals. Hum Brain Mapp, 8:194-208. Linden, D.E., Prvulovic, D., Formisano, E., Vollinger, M., Zanella, F.E., Goebel, R., Dierks, T. (1999) The functional neuroanatomy of target detection: an fMRI study of visual and auditory oddball tasks. Cereb Cortex, 9:815-23. Lo, C.Y., Su, T.W., Huang, C.C., Hung, C.C., Chen, W.L., Lan, T.H., Lin, C.P., Bullmore, E.T. (2015) Randomization and resilience of brain functional networks as systems-level endophenotypes of schizophrenia. Proceedings of the National Academy of Sciences of the United States of America, 112:9123-8. Lubeiro, A., Rueda, C., Hernandez, J.A., Sanz, J., Sarramea, F., Molina, V. (2016) Identification of two clusters within schizophrenia with different structural, functional and clinical characteristics. Prog Neuropsychopharmacol Biol Psychiatry, 64:79-86. MRI and EEG networks in schizophrenia Gomez Pilar et al. 21 Ma, S., Calhoun, V.D., Eichele, T., Du, W., Adali, T. (2012) Modulations of functional connectivity in the healthy and schizophrenia groups during task and rest. Neuroimage, 62:1694-704. Martin-Santiago, O., Gomez-Pilar, J., Lubeiro, A., Ayuso, M., Poza, J., Hornero, R., Fernandez, M., de Azua, S.R., Valcarcel, C., Molina, V. (2016) Modulation of brain network parameters associated with subclinical psychotic symptoms. Prog Neuropsychopharmacol Biol Psychiatry, 66:54-62. 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, 87:60-6. Molina, V., Lubeiro, A., Soto, O., Rodriguez, M., Alvarez, A., Hernandez, R., de Luis-Garcia, R. (2017) Alterations in prefrontal connectivity in schizophrenia assessed using diffusion magnetic resonance imaging. Prog Neuropsychopharmacol Biol Psychiatry, 76:107-115. Nunez, P., Poza, J., Bachiller, A., Gomez-Pilar, J., Lubeiro, A., Molina, V., Hornero, R. (2017) Exploring non-stationarity patterns in schizophrenia: neural reorganization abnormalities in the alpha band. J Neural Eng, 14:046001. Patel, S., Mahon, K., Wellington, R., Zhang, J., Chaplin, W., Szeszko, P.R. (2011) A meta-analysis of diffusion tensor imaging studies of the corpus callosum in schizophrenia. Schizophr Res, 129:149-55. Patenaude, B., Smith, S.M., Kennedy, D.N., Jenkinson, M. (2011) A Bayesian model of shape and appearance for subcortical brain segmentation. Neuroimage, 56:907-22. Rubinov, M., Knock, S.A., Stam, C.J., Micheloyannis, S., Harris, A.W., Williams, L.M., Breakspear, M. (2009) Small-world properties of nonlinear brain activity in schizophrenia. Hum Brain Mapp, 30:403-16. Rubinov, M., Sporns, O. (2010) Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52:1059-69. Salvador, R., Pena, A., Menon, D.K., Carpenter, T.A., Pickard, J.D., Bullmore, E.T. (2005) Formal characterization and extension of the linearized diffusion tensor model. Hum Brain Mapp, 24:144-55. Schoffelen, J.M., Gross, J. (2009) Source connectivity analysis with MEG and EEG. Hum Brain Mapp, 30:1857-65. Segarra, N., Bernardo, M., Gutierrez, F., Justicia, A., Fernadez-Egea, E., Allas, M., Safont, G., Contreras, F., Gascon, J., Soler-Insa, P.A., Menchon, J.M., Junque, C., Keefe, R.S. (2011) Spanish validation of the Brief Assessment in Cognition in Schizophrenia (BACS) in patients with schizophrenia and healthy controls. Eur Psychiatry, 26:69-73. Shim, M., Kim, D.W., Lee, S.H., Im, C.H. (2014) Disruptions in small-world cortical functional connectivity network during an auditory oddball paradigm task in patients with schizophrenia. Schizophr Res, 156:197-203. Smith, S.M. (2002) Fast robust automated brain extraction. Hum Brain Mapp, 17:143-55. Spencer, K.M., Nestor, P.G., Niznikiewicz, M.A., Salisbury, D.F., Shenton, M.E., McCarley, R.W. (2003) Abnormal neural synchrony in schizophrenia. J Neurosci, 23:7407-11. Stam, C.J., de Haan, W., Daffertshofer, A., Jones, B.F., Manshanden, I., van Cappellen van Walsum, A.M., Montez, T., Verbunt, J.P., de Munck, J.C., van Dijk, B.W., Berendse, H.W., Scheltens, P. (2009) Graph theoretical analysis of magnetoencephalographic functional connectivity in Alzheimer's disease. Brain, 132:213-24. Sui, J., Yu, Q., He, H., Pearlson, G.D., Calhoun, V.D. (2012) A selective review of multimodal fusion methods in schizophrenia. Front Hum Neurosci, 6:27. Sun, H., Lui, S., Yao, L., Deng, W., Xiao, Y., Zhang, W., Huang, X., Hu, J., Bi, F., Li, T., Sweeney, J.A., Gong, Q. (2015) Two Patterns of White Matter Abnormalities in Medication-Naive Patients With First-Episode Schizophrenia Revealed by Diffusion Tensor Imaging and Cluster Analysis. JAMA Psychiatry, 72:678-86. Thakkar, K.N., Rosler, L., Wijnen, J.P., Boer, V.O., Klomp, D.W., Cahn, W., Kahn, R.S., Neggers, S.F. (2017) 7T Proton Magnetic Resonance Spectroscopy of Gamma-Aminobutyric Acid, MRI and EEG networks in schizophrenia Gomez Pilar et al. 22 Glutamate, and Glutamine Reveals Altered Concentrations in Patients With Schizophrenia and Healthy Siblings. Biol Psychiatry, 81:525-535. Tournier, J.D., Calamante, F., Connelly, A. (2007) Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution. Neuroimage, 35:1459-72. Van de Steen, F., Faes, L., Karahan, E., Songsiri, J., Valdes-Sosa, P.A., Marinazzo, D. (2016) Critical Comments on EEG Sensor Space Dynamical Connectivity Analysis. Brain Topogr. van den Heuvel, M.P., Mandl, R.C., Stam, C.J., Kahn, R.S., Hulshoff Pol, H.E. (2010) Aberrant frontal and temporal complex network structure in schizophrenia: a graph theoretical analysis. The Journal of neuroscience : the official journal of the Society for Neuroscience, 30:15915-26. van den Heuvel, M.P., Sporns, O., Collin, G., Scheewe, T., Mandl, R.C., Cahn, W., Goni, J., Hulshoff Pol, H.E., Kahn, R.S. (2013) Abnormal rich club organization and functional brain dynamics in schizophrenia. JAMA Psychiatry, 70:783-92. van Diessen, E., Numan, T., van Dellen, E., van der Kooi, A.W., Boersma, M., Hofman, D., van Lutterveld, R. (2015) Opportunities and Methodological Challenges in EEG and MEG Resting State Functional Brain Network Research. Clinical Neurophysiology 126:1468- 81. Varela, F., Lachaux, J.P., Rodriguez, E., Martinerie, J. (2001) The brainweb: phase synchronization and large-scale integration. Nat Rev Neurosci, 2:229-39. von Stein, A., Chiang, C., Konig, P. (2000) Top-down processing mediated by interareal synchronization. Proc Natl Acad Sci U S A, 97:14748-53. Wheeler, A.L., Wessa, M., Szeszko, P.R., Foussias, G., Chakravarty, M.M., Lerch, J.P., DeRosse, P., Remington, G., Mulsant, B.H., Linke, J., Malhotra, A.K., Voineskos, A.N. (2015) Further neuroimaging evidence for the deficit subtype of schizophrenia: a cortical connectomics analysis. JAMA Psychiatry, 72:446-55. Yu, Q., Sui, J., Rachakonda, S., He, H., Gruner, W., Pearlson, G., Kiehl, K.A., Calhoun, V.D. (2011) Altered topological properties of functional network connectivity in schizophrenia during resting state: a small-world brain network study. PLoS One, 6:e25423. Zhang, Y., Brady, M., Smith, S. (2001) Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging, 20:45-57. Table 1: Demographic, clinical and cognitive data in patients and controls. Significant differences with respect to controls are shown for patients * p<0.05; **p<0.001; ***p<0.001. Patients Controls Schizophrenia (EEG, n=39) Schizophrenia (EEG+dMR; n=33) Controls (EEG; n=78) Controls (dMR+EEG; n=27) Age 33.053(8.801) 33.059(8.951) 30.948 (10.839) 34.668 (11.150) Sex (M:F) 23:16 19:14 46:32 18:9 CPZ equivalents (mg/d) 377.901(196.934) 374.802(193.419) NA Duration(months) 95.169(117.388) 83.86(117.456) NA Education years 14.191(3.600) 14.882(3.051) 16.561(2.254) 17.427(2.866) MRI and EEG networks in schizophrenia Gomez Pilar et al. 23 PANSS Positive symptoms 11.702(3.427) 11.388 (3.457) NA PANSS Negative symptoms 17.571(7.309) 15.450 (5.057) NA Total symptoms 53.810(18.892) 53.313 (18.913) NA Total IQ 91.061(14.528) *** 94.701(11.789) *** 113.209(11.088) 109.458 (12.165) Verbal memory 34.262(12.889) *** 35.315(12.345)*** 51.115(8.194) 53.000(7.274) Working memory 16.151(5.010) *** 17.074(4.148) *** 21.626(3.621) 23.140(2.723) Motor speed 58.879(13.781) *** 62.538(12.041) *** 72.610(16.583) 85.503(8.154) Verbal fluency 18.352(5.730) *** 19.613(4.799) *** 27.856(5.155) 28.827(5.177) Proccessing speed 43.700(15.360) *** 45.641(14.672) *** 69.588(14.378) 69.251(14.841) Problem solving 15.253(4.622) 16.317(3.418) 17.524(2.571) 17.042(2.641) WCST perseverative errors (%) 17.921(10.123) *** 21.152 (17.077) *** 9.801(5.141) 8.221(3.573) WCST completed categories 4.419(1.878) *** 4.812 (1.711) ** 5.847(0.610) 5.879 (0.478) MRI and EEG networks in schizophrenia Gomez Pilar et al. 24 Table 2. Structural and functional graph-theory parameters for patient and controls. Mean values are shown with the corresponding sd per group. Task-related modulation is defined as the difference for the corresponding functional parameter between its value at the response and pre-stimulus windows. Structural (dMRI) network Functional (EEG) network Global band Theta band Schizophrenia (n=33) Controls (n=27) Cohen´s d Schizophrenia (n=39) Controls (n=78) Cohen´s d Schizophrenia (n=39) Controls (n=78) Cohen´s d CLC 0.995 (0.002) 0.995 (0.001) Pre-stimulus Modulation 1.006 (0.004) 0.001 (0.001)# 1.007 (0.005) 0.001 (0.002)## 1.009 (0.006) -8.526E-05 (0.004) 1.008(0.004) 0.001 (0.004)# PL 1.018(0.009)* 1.013 (0.005) 0.686 Pre-stimulus Modulation 1.087 (0.028) 0.003 (0.008)# 1.088 (0.027) 0.004 (0.013)## 1.104(0.027) -0.001 (0.020)* 1.099 (0.022) 0.009 (0.025)# -0.432 D 0.329 (0.032) 0.346 (0.030) Pre-stimulus Modulation 0.326 (0.078)* 0.001 (0.009) 0.297 (0.053) -0.001 (0.013) 0.434 0.361 (0.043)* 0.008 (0.025)# 0.3381 (0.042) 0.020 (0.032)# 0.514 * Statistically significant differences between patients and controls (p<0.05) ; # statistically significant within-group task-related modulation(response minus pre-stimulus; p<0.05; ## idem p<0.001. MRI and EEG networks in schizophrenia Gomez Pilar et al. 25 Figure legends Figure 1. Processing pipeline yielding fractional anisotropy values to be used in graphtheory calculations. Figure 2. (A) The 29 EEG channel labels superposed on the structural ROIs. EEG nodes (filled in white) were used to generate functional connectivity matrices from PLV values between each pair of electrodes. The figure illustrates the approximate placement of the EEG electrodes over the ROIs. The list of the 29 electrodes used in the study according to international 10-10 system is: Fp1, Fp2, F7, F3, Fz, F4, F8, FC5, FC1, FCz, FC2, FC6, T7, C3, Cz, C4, T8, CP5, CP1, CP2, CP6, P7, P3, Pz, P4, P8, O1, Oz and O2. Figure 2. (B) Schematic depiction (axial and sagittal views) of the relevant tracts (streamlines) from which fractional anisotropy (FA) was calculated to generate structural connectivity matrices. Streamlines were calculated between each pair of the 84 nodes corresponding to the cortical segmentation are shown as spheres (their sizes are proportional to the actual size of the corresponding ROI). For the sake of clarity, only tracts linking PFC with anterior cingulate, superior temporal, insular and superior parietal cortices and hippocampus and caudate are drawn. Figure 3. Error bars corresponding to the graph-theory parameters with statistically significant differences between patients and controls (from left to right, structural PL, functional pre-stimulus D at the global and at the theta bands, and functional PL taskrelated modulation at the theta band). Circles represent the mean value, while bars indicate the interval of confidence (95%). Figure 4. Scatterplots showing the association between (a) factor scores resulting from the principal components analysis of structural graph-theory parameters (X axis) and scores of the second factor resulting from the principal components analysis of functional graph-theory parameters in the global band (modulation; Y axis); (b) factor scores for the first factor from the principal components analysis of functional graphtheory parameters at the theta band (modulation) and factor scores from the principal components analysis summarizing cognitive scores; (c) positive PANSS scores and global band PL and CLC task-related modulation and (d) structural network (right) Solid dots represent chronic patients, open dots represent FE patients.