Artifacts Removal Techniques for the European iEEG Dataset

In this paper, we propose preprocessing techniques to remove artifacts from Intracranial Electroencephalog-raphy (iEEG) data, including a proposed patient-specific rereferencing method to remove the common noise. The proposed re-referencing method selects the reference electrode based on the electrode placement distribution for each patient. To investigate the effect of the reference electrode location on the obtained results, the proposed re-referencing method is compared to the common average reference method. The proposed preprocessing techniques are evaluated and compared on four patients from the European iEEG dataset. Simulation results indicate that the epilepsy prediction accuracy using an LSTM model based on the proposed patient-specific method has a better performance with an average AUC of 0.83 compared to the same model when using the common average reference method with an average AUC of 0.81.