Latency study of non-convulsive seizures

This paper proposes an algorithm to detect the onset in case of non-convulsive seizures. In the literature no work has been reported in this direction. The algorithm was tested on the scalp EEG database of 5 patients, collected at All India Institute of Medical Science (AIIMS), New Delhi. The database consisted of 13 seizures of different duration. Since EEG signal is random in nature, it was converted into small windows to change it into pseudo stationary data. Time domain analysis was then performed on those small windows and features were extracted. Classification was performed using linear classifier. The obtained sensitivity, specificity and latency using this method were 90.20%, 92.43% and 0.78 seconds respectively.