Rest EEG Hidden Dynamics as a Discriminant for Brain Tumour Classification

Abstract

The hard problem of brain tumour detection is investigated based on rest EEG analysis, trying to ascertain whether the EEG signal contains more hidden useful information than what is clinically employed. A nonlinear analysis of the hidden dynamics is applied to the pair (F3, F4) of EEG leads - describing the electrical activity of the frontal part of the left and right brain hemisphere respectively - in order to detect the possible different features of meningeoma, malignant glioma, and intact brain. The key idea is that underlying systems with different structures produce observed variables with different hidden dynamics. The hidden dynamics of the pair (F3, F4) is tested against a hierarchy of null hypotheses corresponding to nonlinear Markov processes of increasing order. The conditional probabilities of the transition states of the Markov models are represented as sums of Gaussian distributions, whose parameters are estimated by means of Multi-Layer Perceptrons. The minimum order of the accepted Markov models gives an indication of the organisation degree of the signal’s hidden dynamics. A very structured dynamic is detected in both leads (F3, F4) of normal EEGs, confirming the very complex structure of the underlying system. Different correlations between the two hemispheres’ activities seem to discriminate meningeoma, malignant glioma, and no pathological status, while loss of structure can represent a good hint for glioma/meningeoma localisation.