EEG neuro-image reconstruction based on conditional chaos
Saeid Sanei · 2002
Projection of chaotic behaviour into neuro-images for long-term monitoring of EEGs is proposed. The EEG signals are first restored using a block-based blind source separation (BSS) technique. The wavelet packet transform (WPT) of the resulting signals are computed. Then, joint probability density functions (pdfs) of the sequential frames of each signal subject to the values of the neighboring signals, are calculated. Posterior probabilities are computed and time-varying conditional entropies are evaluated. The entropy amplitudes are then mapped into two-dimensional images. Finally, the points are extrapolated to 2D planes and pseudo colored. The reconstructed images are shown to provide valuable information for diagnosis of arrhythmic brain disease. (5 pages)