Prediction of Temporal Lobe Seizures Using the Singular Spectrum Analysis

At present, one goal of the seizure predictions is to use the linear methods to make the prediction simple. In the paper, a linear method called the singular spectrum analysis (SSA) is employed to the prediction of the seizures onset based on the scalp EEG recordings from the epilepsy patients whose focus are in the temporal lobe as well as from the healthy humans. Different from other prediction methods, it doesn't need large scale data and complex algorithm to make it beneficial to the clinical practice. According to the computing experience, about 4 seconds data is enough to make the prediction more efficient and more convenient. In order to evaluate the method, a radial basis function (RBF) neural network model is used to the classification effectively. It is concluded that the healthy people's SSA decreases rapidly and has a 'platform' in the end, but the epileptic patient's SSA decreases gradually, no obvious 'platform' occurs in the end. It is possible for the phenomenon to be available in the temporal lobe seizure predictions.