EEG Channel Selection for Epileptic Seizure Prediction

Early prediction of epileptic seizures could help patients get the appropriate treatment in time and prevent their risks of injury by avoiding dangerous activities. Automatic seizure prediction could help alert the patient to prepare accordingly. For this reason, seizure prediction has been a topic of research for a few decades and various automated methods have been developed towards that direction. However, digital processing of EEG signals captured in numerous frequency channels can be complex with high computational costs. Moreover, not all channels contribute to the overall prediction of epileptic seizures. In order to develop patient-friendly portable devices able to notify patients in real-time of the arrival of an episode, the need to reduce the number of used electrodes is imperative. This review work focuses on the appropriate selection of EEG channels for automated seizure prediction based on the recent literature. Several channel combinations are examined, comparing their prediction performance. Moreover, potential channel combinations to minimize computational costs without losing in prediction accuracy, are highlighted, aiming to provide guidelines for future research in the field.