| Epilepsy is the brain neurons sudden abnormal discharging, which leads to a chronic disease and syndrome of short brain function obstacle, and give patients a great deal of pain. If seizure can be predicted in before going on, the medical staff or relatives of patients can take effective prevention protection for patients. In addition, the prediction of epilepsy research will also help explore the epilepsy mechanism and technology of epilepsy diagnosis and treatment. So there are many scholars devoted themselves to the prediction of the seizures research.The pathogenic factors of seizure are very complex, and mechanism has been not completely interpreted. Therefore, although the seizure prediction algorithm has get development and progress, there are also many problems. For example, accumulated energy curve did not have apparent change in preictal; in the long EEG observation, the relationship between the correlation dimension down and the time length of prediction did not appear significant difference in interictal and preictal; the lyapunov index algorithm can have prediction ability only at very low noise condition, but there is a big noise when EEG is collected, so its reliability reduced greatly.Spike is the most basic paroxysmal abnormal brain electrical activity. In the clinical EEG check, the most important is to identify the presence of spike and sharp. By analysis the epilepsy EEG data, we can found that feature wave like spike,polyspike complex and spike and slow wave complex increase. On the basis of this, this paper analyses the spike rate changes rule in the preictal and proposed a new seizure prediction algorithm based on spike rate.This method analyses the spike rate of six channel of long-term intracranial EEG recording from21cases. First, the high frequency artifacts of EEG were removed SR (Spike Rate) of every segment of EEG. Finally, SRm (mean Spike Rate) is got by smoothing the spike rate using the average filter and the rise phenomenon of SRm in the preictal is judged the seizure basis. Analysis find that,68seizures of87seizures in total from21patient are correctly predicted,, and the sensitivity reached78%with a false prediction rate of0.08/h.By comparing this algorithm with increments of accumulated energy,dynamical similarity index and the phase synchronization based on the wavelet transform, we can get the conclusion:the prediction algorithm based on the analysis of spike rate is small amount of calculation, operation time is short, strong applicability and high reliability. |
