Characterization of Neural Activity Using Complex Network Theory

The current Doctoral Thesis focuses on characterizing brain network dynamics by means of Complex Network Theory to elucidate neural substrates in schizophrenia disorder. Electroencephalographic (EEG) signals, acquired during a cognitive task, were used to obtain connectivity matrices that describe the functional brain network. Graph measures were computed from these matrices using coherence and phase-based measures. These investigations have led to results which have been published, or accepted for publication, in journals indexed in the Journal Citation Reports from Thomson Reuters Web of Science^™(JCR-WOS). Specifically, up to five papers were published between July 2015 and August 2018. Additionally, one more paper was accepted for publication (April 2018). This scientific productivity has allowed writing this work as a compendium of publications. The thematic consistency of the papers included in the Thesis is justified in this introductory chapter (Sect. 1.1). The general context of Biomedical Engineering and neural signal processing is briefly described in Sect. 1.2. Section 1.3 is devoted to schizophrenia disorder. Section 1.4 is oriented to explain physiological underpinnings of the EEG recordings. In Sect. 1.5, the basis of neural oscillations and their generation is explained. Section 1.6 is focused on Event-Related Potential (ERP) and its usefulness in research. Finally, Sect. 1.7 provides the basis for understanding the current tendency to model brain interactions as a graph. The latter, indeed, motivates the research problem and, subsequently, the research questions.

The neural mechanisms of the brain and, particularly, the neural networks involved in pathological behavior as in the schizophrenia disorder, have become a relevant research topic. Hence, the proposal developed in this Doctoral Thesis is focused on analyzing EEG recordings to characterize the underlying neural mechanisms of the brain aimed at finding altered neural substrates in schizophrenia.

In this compendium of publications, different methods and databases were used. To avoid redundancy, a brief summary of the methods followed and the databases used is provided in this chapter. However, a detailed explanation of them is described in the papers of the compendium of publications (see the “Introduction” chapter and the bibliography for further details).

This chapter summarizes the most relevant results found during this Thesis. They have been split according to the different hypotheses of the “Hypotheses and Objectives” chapter, which has an almost direct correspondence with the papers published during this Thesis (see the Introduction chapter and the bibliography for further details).

In this Doctoral Thesis, the characterization of the altered functional network structure and neural dynamics in schizophrenia during a cognitive task has been addressed. Firstly, new evidences for the dysconnectivity hypothesis and the aberrant salience hypothesis in schizophrenia were found. Abnormal response to novel and relevant stimulus (aberrant salience) was recurrently found in patients with schizophrenia in this Thesis. This abnormal response was accompanied by a diminished integration among brain regions connected by long-range interactions. Secondly, novel findings about possible underpinnings of these anomalies were provided by means of the study of the structural network and functional local measures based on regularity. Importantly, a hyperactivation focused on segregated assemblies of neuronal entities during the stimulus expectation can suppose the main difference between the odd response during cognition in schizophrenia. Thirdly, a novel graph measure of network complexity was developed. SCG provides an estimation of the ratio between the order of the network and the amount of information stored in it. The measure is insensitive to changes in connectivity strength and network size for networks large enough (\(N > 30\)). Having shown its virtues, SCG provides new clues about brain network dynamics in schizophrenia during cognition. Finally, the previously mentioned findings using SGC allow us to propose a novel network modeling to describe the brain network dynamics and the main differences between healthy and schizophrenia subjects. This network modeling identifies different reorganization strategies of the brain network as response to an oddball task for patients with schizophrenia, which could be the basis for new studies focused on the heterogeneity in this disorder.

All the studies of this Doctoral Thesis share a common thread: the use of analyses based on graph-theory for enhancing the established knowledge about dynamical connectivity properties of noticeable neural assemblies and their abnormalities in schizophrenia. These studies are intended to be a starting point for a future breakthrough in the study of schizophrenia, in which subgroups inside of this disorder with unique and particular biological characteristics will be identified. The heterogeneity in schizophrenia is a reiterative finding in several studies. This may be a reason for obtaining, sometimes, contradictory results and hindering the replication of results with different databases.