Characterization of neural activity using complex network theory. Application to the identification of the altered neural substrates in schizophrenia

  1. Gómez Pilar, Javier
Supervised by:
  1. Roberto Hornero Sánchez Director
  2. Jesús Poza Crespo Co-director

Defence university: Universidad de Valladolid

Fecha de defensa: 11 December 2018

  1. Pablo Laguna Lasaosa Chair
  2. María García Gadañón Secretary
  3. Daniel Abásolo Baz Committee member
  1. Signal and Communications Theory and Telematic Engineering

Type: Thesis


Neuroscience is a field in constant evolution. From Ramon y Cajal’s works, we recognize neuronal synapses as the basis for brain communication and coordination. These electric connections make up a tangle of pathways of extraordinary density. Until recently, the sophisticated techniques needed for acquiring these complex interactions, mapping their sources and analyzing them with advanced mathematical tools were out of our reach. It is now the right moment in which the technology and the state of the research provide us an opportunity to disentangle the underpinnings of the brain. In this context, this Doctoral Thesis is focused on the study, development and assessment of a specific framework to investigate neural interactions from the perspective of Complex Network Theory. Particularly, the new measures and the novel models provided in this Thesis are aimed at improving our current knowledge of schizophrenia disorder. Schizophrenia is a disabling mental disorder characterized by disintegration of the process of thinking, contact with reality and emotional responsiveness. People with a large variety of symptoms can currently be considered patients with schizophrenia. It is not clear that they all share a common physiological substrate though. The heterogeneity of this illness could be the problem to find the adequate treatment for each one. This problem is accentuated since its prevalence is around 0.5% and patients have a 20% decrease in life expectancy compared with the general population. Genetics, child abuse or cannabis consumption are some of the risk factors of schizophrenia. Nevertheless, not much is known about brain coordination and communication in this disorder compared to general population. As seen throughout this Thesis, Complex Network Theory can significantly contribute in this regard. Neural oscillations can be considered as the main contributing mechanism for enabling coordinated activity during normal brain functioning. Alterations in neural oscillations have been observed in schizophrenia using electroencephalographic (EEG) recordings. Aberrant relevance assignment to stimulus is a common finding in several studies, which could be assessed using event-related potentials (ERPs). Schizophrenia has been also identified as a dysconnectivity syndrome; a concept related to the reduced capacity to integrate information among different brain regions. The concept of connectome and the use of graph theory to model, estimate and simulate the topology and dynamics of the brain network reduce the complexity of this problem with a remarkably flexible representation. Despite the growing use of graph theory in neuroscience in the last decade, only a handful of studies focused on brain dynamics using graphs (chronnectomic) and, to the best of our knowledge, the studies presented here are among the first ones in using this approach for assessing brain dynamics in schizophrenia. All the studies that comprise this Doctoral Dissertation use the neuroelectrical signal from EEG recordings during an auditory oddball task to assess the schizophrenia underpinnings. In particular, a 3-stimulus auditory-oddball paradigm was used for examining cognitive processing as response to both relevant and irrelevantstimuli. Two incremental databases were used in the studies: the former reaching a total of 48 patients with schizophrenia and 87 healthy controls with 16 electrodes; the latter including 39 patients and 78 healthy controls with 32 electrodes. Using local and network measures, changes of the brain activation from prestimulus to cognitive response were characterized. The main findings in this Thesis include: (i) a hyper-segregated state in schizophrenia during the expectation of stimulation; (ii) a deficit in the change from pre-stimulus to response activity in local and network features linked to the previously mentioned hypersegregated state; (iii) a non-remarkable correlation between structural connectivity (obtained by diffusion magnetic resonance imaging, dMRI) and functional connectivity (EEG-based); and (iv) a noticeable reinforcement in the secondary pathways, i.e., pathways weakly connected during pre-stimulus, followed by most of the controls, but the patients’s behavior is split between primary and secondary pathways reinforcement. Additionally, two genuine novelties were introduced: (i) a new measure of graph complexity based on the balance of the edge weights that does not need surrogate data, and (ii) a dynamical network model for the assessment of changes in the synchronization of brain regions during cognition, useful both for healthy and diseased brain interactions. Our findings are consistent with previous reports in schizophrenia, revealing an abnormal salience assignment and disorganized internal representation, as well as desynchronized coupling in long-range brain interactions. On the other hand, the studies of this compendium of publications show a hyper-activation of the prestimulus interval accompanied by reduced entropy followed by a deficit of change during cognition. This highlights that the higher the range of the available frequencies during pre-stimulus (spectral entropy), the higher flexibility or capability to change during the subsequent stimulus, providing a novel link between local and network features, in which patients with schizophrenia show abnormalities. The lack of findings related to the association between structural and functional connectivity, as well as the mentioned division in the behavior of the network dynamics during cognition in schizophrenia, supports the likely heterogeneity in schizophrenia disorder. Further studies must be addressed to corroborate this hypothesis, which could involve a breakthrough in this field. In summary, this Doctoral Thesis evaluates dynamical network features in schizophrenia during an auditory cognitive task, resulting in a reliable characterization of dynamical neural patterns. For this purpose, local measures based on the spectral distribution from the EEG, as well as network characteristics were estimated. Novel network measures and a new network modeling during cognition were presented. The findings of this Thesis reinforce previous works, whereas open the door to new frameworks to study the heterogeneity of schizophrenia.