NEURAL-NETWORK MODELING OF HUMAN ELECTROENCEPHALOGRAM PATTERNS

A general regression neural network model with different processing el ements (PE) fits the observed electroencephalogram signals (EEG) of th e human brain during awake, sleep and rapid eye movement (REM) stages, The number of PEs required for simulation could be an indication of t he complexity of the EEG pattern, About 6.5 times more PEs are require d to simulate the signal during alert eyes open (beta waves) than aler t eyes closed (alpha waves) stage, Also, half the number of PEs are re quired to simulate sleep stage 4 when compared to simulating sleep sta ge 2. REM sleep requires more number of PEs to simulate than sleep sta ge 4, PEs required to simulate short duration 'petit mal' epileptic se izure less than those required to simulate beta waves, but more than t hose required to simulate REM state, In most cases a correlation exist s between the number of PEs and the fractal dimension.