Patient Specific Seizure Onset-Offset Latency Detection using Long- term EEG Signals

摘要

Brain disorders like epilepsy affects the non-seizure life of epileptic patients enormously during seizure attack. EEG signals could be used for diagnosing seizures in epileptic patients. In this work, a patient specific algorithmic using wavelet based features of EEG signals is developed for seizure onset-offset latency detection. Statistical features namely entropy, mean and energy were computed over wavelet subbands. Linear classifier was employed for classification of nonseizure EEG and seizure EEG. The presented algorithmic was tested using long duration continuous EEG of 206 hours from an open source EEG dataset of CHB-MIT. The developed algorithmic obtained average specificity, accuracy, sensitivity onset and offset latency of 98.64, 98.60, 96.43%, 1.7 and 0.9 seconds respectively. These results of present study were also compared with the available techniques of seizure detection reported using same database. The proposed technique showed improvement in seizure onset and offset latency detection comparison to other reported results.