Patient-Specific Remaining Time Forecasting of Epileptic Seizures: Optimal Window and Step Size Approach

An epileptic seizure is a short-term event that occurs due to a sudden, abnormal change in the normal electrical activity of the brain. These changes may manifest with symptoms such as loss of consciousness and involuntary movements. This disease, which has no definitive solution, requires patients to take their medications or move to a safe area in order to experience it more easily. As a contribution to the literature, we propose a model that predicts the remaining time (RT) for the next epileptic seizure and thus provides patients with the temporal opportunity to take these measures. Additionally, we suggest that the most appropriate window (W) and step size (S) values to accurately predict RT should be determined based on patient-based seizure characteristics, unlike existing studies. We tested the proposed model on the CHB-MIT Scalp EEG dataset, which is the benchmark in this field. The results show that the proposed model predicts RT with a high prediction success rate of up to 0.93 R2 and 0.07 MSE. Finally, the conclusion is that it is vital for the patient to determine the W and S values based on the patient's seizure characteristics, rather than using fixed values.