Elsevier

Schizophrenia Research

Volume 201, November 2018, Pages 120-129
Schizophrenia Research

Altered predictive capability of the brain network EEG model in schizophrenia during cognition

https://doi.org/10.1016/j.schres.2018.04.043Get rights and content

Abstract

The study of the mechanisms involved in cognition is of paramount importance for the understanding of the neurobiological substrates in psychiatric disorders. Hence, this research is aimed at exploring the brain network dynamics during a cognitive task. Specifically, we analyze the predictive capability of the pre-stimulus theta activity to ascertain the functional brain dynamics during cognition in both healthy and schizophrenia subjects. Firstly, EEG recordings were acquired during a three-tone oddball task from fifty-one healthy subjects and thirty-five schizophrenia patients. Secondly, phase-based coupling measures were used to generate the time-varying functional network for each subject. Finally, pre-stimulus network connections were iteratively modified according to different models of network reorganization. This adjustment was applied by minimizing the prediction error through recurrent iterations, following the predictive coding approach. Both controls and schizophrenia patients follow a reinforcement of the secondary neural pathways (i.e., pathways between cortical brain regions weakly connected during pre-stimulus) for most of the subjects, though the ratio of controls that exhibited this behavior was statistically significant higher than for patients. These findings suggest that schizophrenia is associated with an impaired ability to modify brain network configuration during cognition. Furthermore, we provide direct evidence that the changes in phase-based brain network parameters from pre-stimulus to cognitive response in the theta band are closely related to the performance in important cognitive domains. Our findings not only contribute to the understanding of healthy brain dynamics, but also shed light on the altered predictive neuronal substrates in schizophrenia.

Introduction

It is well-established that disturbed cognition is a core feature of schizophrenia. Schizophrenia patients often exhibit global IQ deficits (Zanelli et al., 2010) and impairments in several cognitive domains, such as semantic memory (Rossell and Batty, 2008), executive function (Simonsen et al., 2011), and sustained attention (Sánchez-Morla et al., 2009), among others (Sheffield and Barch, 2016; Vöhringer et al., 2013). These impairments are likely related to alterations in prefrontal neural network dynamics in schizophrenia (Mukherjee et al., 2016; Poppe et al., 2016). However, the exact relationship between neural network abnormalities and cognitive impairment remains unclear.

Cognition has not only been exhaustively studied using a neuropsychiatric approach both in healthy individuals (Leech and Sharp, 2014) and in schizophrenia patients (Moustafa and Gluck, 2011; Vöhringer et al., 2013), but also from a neuroscientific perspective (Li et al., 2016; van den Heuvel and Fornito, 2014). In this context, a dynamical causal model of the brain behavior has been previously proposed (Friston et al., 2003). Despite the number of virtues of the model, dynamical causal modeling requires a high computational cost and the adjustment of several parameters (Thai et al., 2009). Additionally, the complexity of this model makes it rather difficult to draw direct relationship to brain networks without a strong a priori hypothesis. For these reasons, intuitive models focused on explaining the observed neurodynamics, could be helpful. In this regard, the framework of the predictive coding could be the basis to provide a Bayesian inference of the observed environment (Kilner et al., 2007). Predictive coding is based on minimizing prediction error through recurrent interactions among cortical hierarchy levels (Kilner et al., 2007). The neural activity encoding a particular brain state determines where the current dynamics are within the hierarchical sequence (Friston and Kiebel, 2009). Therefore, the encoding of a particular state would have a predictive capability of the subsequent state. Perceptual alterations could be then explained by abnormalities in the dynamic mechanisms of predictive coding (Hohwy et al., 2008).

In this study, we propose an intuitive and reliable model of neural network dynamics during a cognitive task, in which the error between the modeled network and the real brain network is recurrently minimized. Thus, the brain network during the pre-stimulus activity (i.e., prior to stimulus presentation or perception) determines the brain network during the subsequent state. It is necessary, therefore, to characterize the brain network in different moments of the task. One approach being considered would be to directly compare these network parameters, i.e. an arithmetic difference, which would summarize the brain dynamics. This approach can be useful to characterize the network changes, but not the underlying neural mechanisms of such changes. A probabilistic model is, therefore, required in order to identify the neural underpinnings associated with the cognitive task. For that purpose, graph-theoretical analyses combined with EEG can be used to provide a mathematical representation of the functional brain network for studying rapid changes in the coordination and synchronization between different regions. Based on previous evidence about the importance of rapid changes in the cognitive processing (Varela et al., 2001), EEG becomes a suitable tool to analyze brain network changes in the range of milliseconds, unreachable by other neuroimaging techniques, such as fMRI. In addition, it is crucial the use of complementary network measures to obtain a comprehensive characterization of the functional brain network (Rubinov and Sporns, 2010). It is generally accepted that functional brain network is well-connected (Power et al., 2013) and complex (Liu et al., 2008). Furthermore, it exhibits an optimal balance between integration and segregation (Deco et al., 2015), as well as between regularity and irregularity (Tononi et al., 1998). Abnormalities in the previously mentioned brain network features have been reported in schizophrenia (Liu et al., 2008; van den Heuvel and Fornito, 2014; Yeo et al., 2016). Therefore, a combination of the previous network characteristics should be helpful to characterize brain network dynamics related to cognition in schizophrenia.

Dysfunctional interactions between brain areas have been repeatedly suggested as a relevant contribution to explain the mental alterations in schizophrenia (Bjorkquist et al., 2016; Friston and Frilh, 1995; Whalley, 2005). Within this framework, disrupted connectivity in long-range interactions plays a central role in this disorder (Dickerson et al., 2010; Friston et al., 2016; Gomez-Pilar et al., 2015; Sigurdsson et al., 2010). It is noteworthy that a relationship between long-range interactions and low frequency bands, such as delta and theta, has been proposed (Uhlhaas and Singer, 2010). Therefore, it is not surprising that noticeable findings have been usually reported in the literature about the strong association between schizophrenia and brain connectivity in the low EEG frequency bands (Ford et al., 2002; Koenig et al., 2001; Uhlhaas and Singer, 2010). Alterations on low frequency bands have been related to a temporal misalignment of working memory function in schizophrenia (Kikuchi et al., 2007). In this regard, it was suggested that the neural activity underlying working memory may be abnormally dominated by slow frequencies in schizophrenia (Northoff and Duncan, 2016). Similarly, theta oscillations were proposed to be the basis for memory integration (Buzsáki, 2005) and top-down processing (Uhlhaas et al., 2008), both impaired in schizophrenia patients (Clare, 1993; Rossell and Batty, 2008). In addition, it has been suggested that cognitive control deficits may contribute to episodic memory deficits in schizophrenia (Barch and Sheffield, 2014), in which hippocampal and prefrontal regions could play an important role. This, jointly with our previous studies (Bachiller et al., 2015; Gomez-Pilar et al., 2018c), lead us to claim the importance of theta band to characterize the dynamical cognitive network. The analysis of the electric brain activity at low frequencies during the performance of an oddball task (related to working memory function and top-down processing) could then enhance our understanding of memory mechanisms in schizophrenia.

In the last decade, several studies assessed the brain network changes during a cognitive task in schizophrenia and healthy individuals, some of them by means of an oddball task (Bachiller et al., 2015; Reijneveld, 2011; Shim et al., 2014). They reported differences in connectivity and/or network features during the cognitive processing. However, for the sake of comparability, it would be appropriate to go a step further and identify a cognitive network model to explain the observed neural dynamics. In a previous study (Gomez-Pilar et al., 2018c), we suggested that network differences between a healthy and a schizophrenia brain could be related with secondary pathways (i.e., pathways between nodes weakly connected) of the brain network during the pre-stimulus activity. These pathways would be strongly reinforced during the cognitive processing, while other connections would remain almost unchanged. These differences could be specifically linked to frequency bands related to memory and hippocampal activity (i.e. low frequency bands).

Hence, the present study aimed at elucidating the dynamical network model during a cognitive task that better fits the brain network changes in a healthy population, as well as the possible abnormalities in schizophrenia. To avoid inter-subject variability, we performed an individualized approach that provides a specific network model for each subject.

Section snippets

Study subjects

Thirty-five schizophrenia patients were recruited from the Psychiatry Department at the University Hospital of Valladolid (Spain). Diagnoses were made according to the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-V) criteria (American Psychiatric Association, 2013). Fifty-one healthy control subjects, keeping a non-statistically significant age and gender ratio, were also included in the study. Inclusion/exclusion criteria were undertaken identically as in our

Network dynamics

A visual comparison of the averaged brain networks before and after the stimulus onset (see Fig. 2) shows a global increase of the edge weight values for both groups, though this increase is more noticeable for controls. The brain networks were visualized using the BrainNet Viewer (Xia et al., 2013). To assess network evolution across time, a sliding window approach was used. Windows of 300 ms with an overlap of 90% were selected for network measures computation. Fig. 3A shows the associated

Discussion

To the best of our knowledge, this is the first study that combine network modeling and EEG recordings to determine a model of network dynamics during cognition for healthy and schizophrenia subjects. The proposed network modeling effectively predicts the functional brain network of the cognitive response from the pre-stimulus activity.

Conclusions

We provided direct evidence of the predictive capability of the proposed model to ascertain the functional brain behavior during cognition. Our results support the idea that schizophrenia is associated with significant abnormalities in the relation between neural dynamics during the pre-stimulus and cognitive response, which are directly related to cognitive performance. Furthermore, we presented a new model of network organization during cognition based on graph theory measures, which could be

Role of the funding source

This research project was supported in part by “Ministerio de Economía y Competitividad” and FEDER (TEC2014-53196-R), by ‘European Commission’ (POCTEP 0378_AD_EEGWA_2_P), by ‘Consejería de Educación de la Junta de Castilla y León’ (VA037U16), by “Fondo de Investigaciones Sanitarias (Instituto de Salud Carlos III)” under projects FIS PI11/02203 and PI15/00299, and by “Gerencia Regional de Salud de Castilla y León” under projects GRS 932/A/14 and GRS 1134/A/15. A. Lubeiro was in receipt of a

Contributors

J. Poza, R. Hornero and G. Northoff designed the study. J. Gomez-Pilar, C. Gómez, J. Poza and R. Hornero analyzed the EEGs and undertook the statistical analyses. J. Gomez-Pilar wrote the manuscript and performed the computational modeling. V. Molina recruited the patients and performed clinical assessment. A. Lubeiro and B. Cea-Cañas performed the electroencephalographic recordings and the cognitive assessment. All authors contributed to the article, revised and approved the final manuscript.

Conflict of interest

The authors report no biomedical financial interests or potential conflicts of interest.

Acknowledgement

The authors thank Mr. Martinez-Cagigal for providing the Matlab scripts that significantly enhance the appearance of some figures of the manuscript. We also would like to express our gratitude to the Psychiatry Service of the Clinical University Hospital of Valladolid (Spain), for their help and support and the altruistic volunteers for participation.

References (64)

  • J. Gomez-Pilar et al.

    Deficits of entropy modulation in schizophrenia are predicted by functional connectivity strength in the theta band and structural clustering

    NeuroImage Clin.

    (2018)
  • J. Hohwy et al.

    Predictive coding explains binocular rivalry: an epistemological review

    Cognition

    (2008)
  • M. Kikuchi et al.

    Native EEG and Treatment Effects in Neuroleptic-naïve Schizophrenic patients: Time and Frequency Domain Approaches

    (2007)
  • T. Koenig et al.

    Decreased functional connectivity of EEG theta-frequency activity naïve patients with schizophrenia: preliminary results

    Schizophr. Res.

    (2001)
  • P. Lakatos et al.

    The spectrotemporal filter mechanism of auditory selective attention

    Neuron

    (2013)
  • S. Ma et al.

    Modulations of functional connectivity in the healthy and schizophrenia groups during task and rest

    NeuroImage

    (2012)
  • S. Micheloyannis et al.

    Small-world networks and disturbed functional connectivity in schizophrenia

    Schizophr. Res.

    (2006)
  • B. Moghaddam

    Bringing order to the glutamate chaos in schizophrenia

    Neuron

    (2003)
  • A.A. Moustafa et al.

    Computational cognitive models of prefrontal-striatal-hippocampal interactions in Parkinson's disease and schizophrenia

    Neural Netw.

    (2011)
  • G. Northoff et al.

    How do abnormalities in the brain's spontaneous activity translate into symptoms in schizophrenia? From an overview of resting state activity findings to a proposed spatiotemporal psychopathology

    Prog. Neurobiol.

    (2016)
  • J.D. Power et al.

    Article evidence for hubs in human functional brain networks

    Neuron

    (2013)
  • S. Rossell et al.

    Elucidating semantic disorganisation from a word comprehension task: do patients with schizophrenia and bipolar disorder show differential processing of nouns, verbs and adjectives?

    Schizophr. Res.

    (2008)
  • M. Rubinov et al.

    Complex network measures of brain connectivity: uses and interpretations

    NeuroImage

    (2010)
  • E.M. Sánchez-Morla et al.

    Comparative study of neurocognitive function in euthymic bipolar patients and stabilized schizophrenic patients

    Psychiatry Res.

    (2009)
  • N. Segarra et al.

    Spanish validation of the Brief Assessment in Cognition in Schizophrenia (BACS) in patients with schizophrenia and healthy controls

    Eur. Psychiatry

    (2011)
  • J.M. Sheffield et al.

    Cognition and resting-state functional connectivity in schizophrenia

    Neurosci. Biobehav. Rev.

    (2016)
  • M. Shim et al.

    Disruptions in small-world cortical functional connectivity network during an auditory oddball paradigm task in patients with schizophrenia

    Schizophr. Res.

    (2014)
  • N.J. Thai et al.

    Disconnected brains: what is the role of fMRI in connectivity research?

    Int. J. Psychophysiol.

    (2009)
  • G. Tononi et al.

    Complexity and coherency: integrating information in the brain

    Trends Cogn. Sci.

    (1998)
  • American Psychiatric Association

    Diagnostic and Statistical Manual of Mental Disorders

    (2013)
  • A. Bachiller et al.

    Decreased spectral entropy modulation in patients with schizophrenia during a P300 task

    Eur. Arch. Psychiatry Clin. Neurosci.

    (2014)
  • A. Bachiller et al.

    A comparative study of event-related coupling patterns during an auditory oddball task in schizophrenia

    J. Neural Eng.

    (2015)
  • Cited by (18)

    • Dynamics and synchronization control in schizophrenia for EEG signals

      2022, Biomedical Signal Processing and Control
      Citation Excerpt :

      Schizophrenia patients exhibit connection and activity disorders of the brain networks [50,51]. Consistent with a previous study [52], our results showed that the brain networks had stronger topological connections in the schizophrenia patients than healthy controls. This could be related to the decrease in Cp and Eloc and the smaller number of hubs.

    • Extending the “resting state hypothesis of depression” – dynamics and topography of abnormal rest-task modulation

      2021, Psychiatry Research - Neuroimaging
      Citation Excerpt :

      Using research from healthy subjects and other disorders in addition to MDD, we illustrate the importance of considering both intrinsic and extrinsic activity of the brain, i.e., rest and task states, as interdependent components that permeate amongst one another. The importance of abnormal rest-task modulation for MDD and psychiatric disorders in general is supported by a recent studies showing abnormal rest-task modulation in autism (Lian and Northoff, 2021), anxiety disorders (Tumati et al., 2021), and schizophrenia (Gomez-Pilar et al., 2018; Northoff and Gomez-Pilar, 2021). There is similarity in that all disorders share abnormal spatiotemporal dynamics as a core mediator of modulation between rest and task states.

    View all citing articles on Scopus
    View full text