The Study of Vigilance using Neural Networks Analysis of EEG

This thesis is concerned with the analysis of the mastoid electroencephalogram (EEG) using parametric modelling and neural network techniques in order to assess the vigilance of a human subject. One possible application of this work would be the design of a monitoring system for tracking the transitions within the vigilance continuum. A system for monitoring the vigilance of individuals involved in actions where loss of vigilance might be critical is highly desirable. Such a system should estimate continuously, non-invasively, in real-time and with satisfactory accuracy the fluctuations of the vigilance level. To our knowledge, no vigilance monitor was designed using a single channel of EEG, without the use of additional electrooculographic (EOG) or electromyographic (EMG) information. Moreover, none of the studies involving EEG analysis in general have focused on the mastoid EEG. Although extremely noisy, the mastoid EEG presents significant advantages for on-line, real-life applications: it is not only non-invasive, but also non-obtrusive and very convenient to use. Parametric modelling using autoregressive coefficients has previously been used in EEG analysis, but not in vigilance studies. The strategy adopted in this thesis consists of training neural networks with spectral features extracted from the mastoid EEG. The results are validated against the expert scoring of the vigilance level performed by visual inspection of the central EEG, EOG and EMG signals, and against the results obtained by training similar networks with information extracted from the central EEG (widely recognised as providing useful information for vigilance level assessment). To improve the performance of the neural networks a Kohonen map-based technique for filtering the training data is proposed (this allows labels assigned by an expert to 15-second epochs to be transcripted reliably to one-second segments). The results presented demonstrate conclusively that the tracking of fluctuations from alertness to drowsiness within the vigilance continuum is possible by neural network analysis of a single channel of EEG recorded from the mastoid site.