Epileptic Seizure Prediction Using Random Projections

The prediction of epileptic seizures could provide a significant impact for an estimated 50 million people worldwide. A frequently used tool for brain monitoring is the electroencephalogram, or EEG One potential technique for seizure prediction from EEG data is that of template formation, as first described by Te Van Quyen et al. [9]. However, the formation of such templates can be computationally intensive, and requiring large dimensional measurements. Mitigating the large dimensionality, we utilize random projections upon an EEG signal, and compare the impact of the projected template formation to the template based upon an entire data set. Preliminary results indicate that there is little difference between the between the projected and full templates, and investigate different data selection methods.