An Epileptic Seizure Prediction Algorithm from Scalp EEG Based on Morphological Filter and Kolmogorov Complexity

Epilepsy is the most common neurological disorder in the world, second only to stroke. There are nearly 15 million patients suffer from refractory epilepsy, with no available therapy. Although most seizures are not life threatening, they are an unpredictable source of annoyance and embarrassment, which will result in unconfident and fear. Prediction of epileptic seizures has a profound effect in understanding the mechanism of seizure, improving the rehabilitation possibilities and thereby the quality of life for epilepsy patients. A seizure prediction system can help refractory patients rehabilitate psychologically. In this paper, we introduce an epilepsy seizure prediction algorithm from scalp EEG based on morphological filter and Kolmogorov complexity. Firstly, a complex filter is constructed to remove the artifacts in scalp EEG, in which a morphological filter with optimized structure elements is proposed to eliminate the ocular artifact. Then, the improved Kolmogorov complexity is applied to describe the non-linear dynamic transition of brains. Results show that only the Kolmogorov complexity of electrodes near the epileptogenic focus reduces significantly before seizures. Through the analysis of 7 long-term scalp EEG recordings from 5 epilepsy patients, the average prediction time is 8.5 minutes, the mean sensitivity is 74.0% and specificity is 33.6%.