Characterization of dynamical neural activity by means of EEG dataapplication to schizophrenia

Resumo

Schizophrenia is a disabling, chronic and severe mental illness characterized by disintegration of the process of thinking, contact with reality and emotional responsiveness. Schizophrenia has been related to an aberrant assignment of salience to external objects and internal representations. In addition, schizophrenia has been identified as a dysconnection syndrome, which is associated with a reduced capacity to integrate information among different brain regions. Relevance attribution likely involves diverse cerebral regions and their interconnections. As a consequence, many efforts have been devoted to identifying abnormalities in the cortical connections and their relation to schizophrenia symptoms and cognitive performance. Neural oscillations are one of the largest contributing mechanism for enabling coordinated activity during normal brain functioning. Alterations in neural oscillations and cognitive processing in schizophrenia have long been assessed using electroencephalographic (EEG) recordings (i.e. time-varying voltages on the human scalp generated by the electrical activity on the cerebral cortex). Event-related potentials (ERP) depict EEG data as a response to a cognitive task. ERP analyses are used to gain further insights into the neural mechanisms underlying cognitive dysfunctions. In this Doctoral Thesis, a 3-stimulus auditory-oddball paradigm was used for examining cognitive processing as response to both relevant and irrelevant stimuli. A total of 69 ERP recordings were analyzed in the research papers included in the Thesis, which comprises 20 chronic schizophrenia patients, 11 first episode patients and 38 healthy controls. This Doctoral Thesis is focused on the study, design and application of biomedical signal processing methodologies in order to facilitate the understanding of cognitive processes altered by the schizophrenia. EEG data were examined using a two-level analysis: (i) local activation studies to quantify functional segregation of the brain network, by means of spectral analysis and by assessing neural source generators of P3a and P3b components; and (ii) EEG interactions studies to explore functional integration across brain regions, including pair-wise couplings and exploring hierarchical organization of neural rhythms. Functional segregation aims to identify the brain areas dedicated to specific processing tasks. As a first step, spectral analysis of local activation was performed. Three local activation measures were computed: the relative power (RP) (i.e. the proportion of total power attributable to a given frequency band), the median frequency (MF) (i.e. the frequency which comprises the 50% of the power) and the spectral entropy (SE) (i.e. a measure of the irregularity of the EEG data). RP analyses showed an increase of power from the baseline window to the response window for low frequency bands and a decrease for high frequency bands. Nevertheless, the changes were statistically significantly higher (p < 0:01) in controls than in patients. In addition, MF and SE revealed a widespread decrease from baseline to response window for healthy controls, whereas these changes were lower in schizophrenia patients. Our findings also suggested a statistically significantly larger (p < 0:01) MF and SE decrease as a response to target stimuli than as a response to distractor stimuli. Secondly, source imaging techniques were applied to detect neural generators that contribute to the scalp recorded ERP as a response to target (P3b) and distractor (P3a) tones. Our findings were consistent with previous reports, revealing a lower P3a and P3b source activation mainly in frontal and cingulate regions for schizophrenia patients than for healthy controls. Likewise, the intra-group differences between P3a and P3b were larger in patients than in controls, suggesting an inefficient hyperactivation during the processing of target stimuli. On the other hand, functional integration evaluates the dependencies among brain areas. Functional neural coupling analyzes the statistical dependence between the neural activity at different EEG electrodes. In this study, three complementary functional connectivity measures (wavelet coherence (WC), phase-locking value (PLV) and Euclidean distance (ED)) were applied to analyze correlation, synchrony and similarity patterns. In comparison to healthy controls, schizophrenia patients are characterized by a lack of increase of coupling from baseline to response in the theta band and a lack of decrease for beta2 band. These findings suggested that schizophrenia patients failed to response to relevance (i.e. they are not able to change their connectivity patterns between the auditory response and pre-stimulus baseline). In addition, EEG rhythms in different frequency bands can interact with each other, which reflects the complex and hierarchical organization of cognitive processes. This Doctoral Thesis evaluated event-related phase amplitude coupling (ERPAC), obtaining an association between alpha phase and gamma amplitude. Higher prevalence of alpha-to-gamma ERPAC after stimulus onset was found over central-parietal brain areas than over frontal and temporal brain regions. These findings could evidence the role of alpha rhythms as a core feature of cortical communication. In summary, our proposal evaluates time-frequency EEG data when subjects were performed an auditory cognitive task, obtaining a reliable characterization of dynamical neural patterns. For this purpose, segregation and integration were characterized by means of a two-level analysis of EEG data including: spectral analysis, neural source generators, functional connectivity and hierarchical complex organization of neural rhythms. Our findings revealed that schizophrenia patients showed an attention-dependent modulation of spectral distribution, source generators and functional connectivity in specific frequency bands. In conclusion, these results support the aberrant salience and dysconnection hypotheses: schizophrenia patients show a failure to contextualize stimulus processing through a failure on neuronal firing synchronization.