Scaling behavior in human EEG
Download main paper. (Click on further readings above for more references!)

Abstract: Human resting EEG is usually analyzed via Fourier or chaos analysis which carry assumptions of linearity and low-dimensional attractors, respectively. Complex systems may not exhibit simple structures in these measures, which motivated us to consider scaling analyses. Using the method of detrended fluctuation analysis (DFA), we have shown that the resting EEG commonly exhibits power-law scaling behavior over two temporal ranges. This allows a summarization of the dynamics across temporal scales with just two parameters. Furthermore, scalp EEG consists of 128 data channels, and there is a desire in clinical applications to reduce this data. Yet the fairly low spatial resolution of scalp EEG prohibits spatial scaling analysis in the conventional sense. Instead we consider the values of the two temporal scaling exponents as statistical distributions over the 128 electrodes, and show that the moments of these distributions exhibit their own scaling behavior. This allows an extremely concise quantification of the brain dynamics globally in terms of just one or two parameters. It is shown that these global parameters appear effective at distinguishing normal subjects from those with acute cerebral stroke, suggesting possible utility as a clinical diagnostic tool. New experiments are now being planned to examine scaling behavior during so-called phase transitions in the brain, e.g., spontaneous perceptual shifts when viewing ambiguous figures.

1:30pm -3:00pm, HSE 810.