Computational Intelligence Approaches to Brain Signal Pattern Recognition

Analysis of electrophysiological brain activity has long been considered as one of indispensable tools enabling clinicians and scientists to investigate various aspects of cognitive brain functionality and its underlying neurophysiological structure. The relevance of electroencephalogram (EEG) in particular, due to its inexpensive and most importantly, non-invasive acquisition procedure, has been reflected in the abundance of clinical applications and the diversity of areas of research studies it has contributed to. These studies lie within the realm of brain science understood nowadays in a broad sense embracing and linking interdisciplinary fields of neuroimaging, cognitive psychology and neurophysiology among others. In medical practice, EEG is used more pragmatically to support clinicians in their effort to establish the presence, severity and cerebral distribution of neurological disorders. Epilepsy diagnostic serves as a prime example in this regard (Fisch, 1999). The complex nature of brain signals and the intricacies of the measurement process involved (Fisch, 1999; Niedermeyer & Lopes da Silva, 2004), particularly in the case of EEG, render their analysis and interpretation challenging (Kaiser, 2005). Historically, these signals used to be examined only qualitatively based on routine visual inspection and the experience of responsible technicians or practitioners. With the advent of the era of digital biosignal recordings, computerised quantitative electroencephalography gained notable popularity as a supplementary tool enhancing objectiveness of analysis (Kaiser, 2005). The fast pace of technological advancement, considerable progress in neuroscience and neuroengineering along with growing investments in medical and health sectors among others have opened up new possibilities for automated EEG analysis systems. A continually growing scope for their applications set dominant design trends and imposed requirements regarding their functionality that prevail in today’s practice and research. One of the key points in this regard is the need for the increased independence, autonomy and thus the improved reliability of such systems. This has led to a more comprehensive formulation of a computational problem of brain signal analysis within the realm of pattern recognition, which facilitates a more generic description of existing approaches, and development or reuse of suitable pattern recognition methods. In consequence, the notion of brain signal pattern recognition has been introduced to refer to the underlying concept of processing raw data with the aim of acting upon its category (Niedermeyer & Lopes da Silva, 2004; Duda et al., 2001). The objective is to identify patterns in electrophysiological brain activity that are O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m