Nonlinear Dynamics Analysis Of EEG In Visual Processing

Electroencephalograms (EEGs) record electrical activities that have potential to reflect different states of the human brain. The application of the Nonlinear Dynamics (ND) in the EEG analysis has recently developed and achieved some successes. According to the ND theory, EEGs are nonlinear time series produced by the brain and exhibit complex behavior. Nonlinear parameters, such as the Correlation Dimension (D2), the Approximate Entropy (ApEn), the Largest Lyapunov Exponent (L1), etc., characterize the complexity of ND systems. Most of these ND studies can be divided into two categories: physiological studies and pathological studies. Generally, the former concentrates on the evaluation of conscious and mental states (i.e. classification of different sleep stages, effects of anesthetics, etc.) and the latter fixes attention on the research of neural and psychotic illness (i.e. epilepsy, brain injury, Alzheimer's disease, etc.). Image recognition and categorization are daily tasks for the ordinary people. The underlying visual processing mechanism of the brain has attracted substantial attention lately. Several studies have been conducted on the analysis of EEGs to address dynamic activities of the brain during the visual perception and categorization. The results of Independent Component Analysis (ICA) show that individual independent components might index the neural synchrony within and between the intracranial brain sources. Moreover, the outcomes of improved Event Related Potentials (ERP) analysis demonstrates that the event-related changes in the local field activities might modulate the strength of the spike-based communication between the cortical areas during and after the target recognition. In this paper, we choose the ND method to detect the complexity changes of the EEGs in the visual processing. The dimensional complexity (DCx) values are estimated. A similarity index is constructed to precisely detect the changes. Scalp maps are drawn to visualize the dynamic properties. Significantly lower DCx values are observed at most channels when subjects are performing visual recognition and categorization tasks. This decrease in DCx values may be produced by the neural synchronization of cortical field activities caused by the visual processing. Our results may be helpful to understand the nonlinear human EEG activities in the visual processing.