Low computational complexity EEG epilepsy data classification algorithm for patients with intractable seizures

This paper presents a low computational complexity algorithm for epileptic Electroencephalogram (EEG) data classification. A patient-specific approach is used to identify anomalous data with potential seizure activity. We use a combination of a Finite Impulse Response (FIR) filter to smooth out the signal and a signal thresholding step to determine whether the analyzed data segment is normal or abnormal. The algorithm has been tested on seven subjects each with more than 25 hours of recorded data, resulting in an average sensitivity of 97% and a false positive rate of 0.25 per hour. The proposed algorithm finds applications in the automated support systems for ambulatory patients to reduce storage requirements by eliminating data that is neither in the pre-ictal nor in the post-ictal states. Also, it enables real time data analysis of EEG signals.