Neuro-image Reconstruction based on Bayesian Chaos Evaluation

S. Sanei and W.C. Tan (Singapore)


EEG, Brain map, Bayesian, Chaos


EEG brain map reconstruction and analysis in both time and frequency domains has great attraction due to exploitation of maximum information in those signals. Here, A new imaging modality has been investigated and proposed for long-term monitoring of both rhythmic and arrhythmic EEGs. The signals are restored using Blind Source Separation (BSS) technique. Then second order joint pdf of the sequential frames of each signal is calculated. Posterior probabilities are then computed and time-varying conditional entropies are measured. The entropy amplitudes are then mapped to the two dimensional images. Finally the points are extrapolated to the 2D planes and pseudo colored. Red regions then represent highest entropy and blue areas represent lowest entropy. The method provides sufficient information about brain function. Moreover, it complements the traditional time and frequency based image reconstruction techniques for which either a very long record of the data has to be manipulated or could be used for rhythmic abnormalities only.

Important Links:

Go Back