Cj. Stam et al., INVESTIGATION OF EEG NONLINEARITY IN DEMENTIA AND PARKINSONS-DISEASE, Electroencephalography and clinical neurophysiology, 95(5), 1995, pp. 309-317
Many recent studies based on the surrogate data method failed to ident
ify significant non-linearity in the EEG. in this study we examine whe
ther the use of a different embedding method (spatial instead of time
delay), and calculation of Kolmogorov entropy (K-2) and the largest Ly
apunov exponent (L(1)) in addition to the correlation dimension (D-2),
can distinguish the EEG form linearly filtered noise. We have calcula
ted D-2, L(1) and K-2 of original EEG epochs and surrogate (phase rand
omized) data in 9 control subjects, 9 demented patients and 13 Parkins
on patients. The correlation dimension D-2 and the largest Lyapunov ex
ponent L(1) could distinguish between the EEG tracings and the surroga
te data. Demented patients had significantly lower D-2 and L(1) compar
ed to controls. L(1) was higher in Parkinson patients than in demented
patients. Contrary to other studies that have used the Theiler surrog
ate data method, we find evidence for non-linearity in normal and abno
rmal EEG during the awake/eyes closed state. Apparently it is the spat
ial structure in the EEG that exhibits much of the non-linear structur
e. Furthermore, non-linear EEG measures show more or less specific pat
terns of dysfunction in dementia and Parkinson's disease.