On the Analysis of Seizure Onset in the Eeg: the Application of Constrained Ica

Epileptic seizures are commonly studied using the electroencephalogram (EEG) for the measurement of brain electrical signals, which often show characteristic seizure waveforms. The automatic extraction of this seizure waveform from the recorded data has many implications, including the possibility of a real time system to warn the patient of an impending seizure. Such detection has hitherto proved difficult because of the non-stationarity of seizure waveform and corruption of the EEG signals by artifacts and other waveforms present in the measurements. We demonstrate the use of the recently published constrained independent component analysis (cICA) algorithm for the extraction of seizure waveform from EEG recordings, from time periods prior to the clinical onset of the seizure.