利用高频脑电的局灶性癫痫患者癫痫发作检测

Aiming at the phenomenon that the most existing methods of automatic seizures detection are used to low frequency EEG and ignored the high-frequency components, we attempt to adopt the high frequency component of long term scalp EEG to detect seizures in focal epilepsy patients in this research. Gamma band is extracted in 19 channel EEG using discrete wavelet transform in a sliding window, and the information flow characteristics of each band is evaluated with directional transfer function algorithm. The intensity characteristics of the outgoing information are used to reduce the dimensions. The features are classified by support vector machine (SVM). Five-fold cross validation indicates that the proposed strategy achieves an excellent performance with the average accuracy of 98.4%, the average selectivity of 60.7%, the average sensitivity of 93.4%, the average specificity of 98.4% and the average detection rate of 95.9%, and the gamma band is endowed with higher classification effect. For patients with focal epilepsy, the intensity of the gamma band outflow during seizure attack is significantly concentrated and enhanced in some brain regions. Simultaneously it validates the point that seizure attack is related to high frequency components in the literatures.