AUTOMATIC PROBABILISTIC ANALYSIS OF ABNORMAL ADULT BRAIN SIGNAL

Electroencephalograph (EEG) signal is brain electrical activity and recorded from the scalp by using electroencephalogram. EEG reflects functional status of the brain. EEG signal has a unique contribution and meaningful to recognize the electrical function of the brain. This signal is interpreted visually by electroencephalographers (EEGers) to be used as one of diagnosis instrument for some diseases in the brain. The parameter of visual analysis for awake patients includes frequency, amplitude, and symmetry of background signal, distribution, morphology, and also reactivity. The normality or abnormality of signal could be determined based on these parameters. Unfortunately, the visual analysis has some weaknesses. First, it is the lack of reliability due to the various capability of EEGers. Secondly, if the spatial resolution of EEG increases, EEGers need more time to interpret the signal. Hence, an automatic analysis based on computer was developed to improve the scoring reliability and able to perform a fast calculation process for a large EEG data. In this work, we developed a probabilistic analysis method to justify the abnormality of EEG signal based on the quantitative parameters. This quantitative parameter estimation adopted from the method used by the clinicians for visual reading. However, the result of calculation could not be directly used to determine the abnormality of EEG signal. So, the probabilistic analysis was used with 14 features of the calculation result parameters on each 65 of normal and abnormal EEG data. The performance of the system for normality and abnormality were validated by 10-fold cross validation using discriminant analysis resulted 78.46% of accuracy. In conclusion, it could be said that features extraction adopted from visual analysis are able to distinguish between normal and abnormal in the clinical EEG signals