Epileptic Seizure Prediction Using Geometrical Features Extracted from HRV Signal

Abstract

The prediction of epileptic seizures in patients can help prevent many unwanted risks and excessive suffering. In this research, electrocardiography (ECG) signals for 7 patients under supervision and aged 31–48 years old, both male and female, were recorded. To predict recurrent seizures, we aimed to differentiate heart rate variation (HRV) signals in both seizure and seizure-free states. The evaluation of ECG signals alone was not able to predict recurrent seizures. Furthermore, we extracted geometrical features from our recordings. The results showed distinct differences between the two states with a high percentage of performance evaluation. Therefore, the proposed algorithm was successful and can be used to predict these epileptic seizures, thus decreasing the pain and suffering in patients. In this study, we have presented a new approach based on the geometric feature from extracted HRV analysis. We achieved a sensitivity of 100%, an accuracy of 90%, and a specificity of 88.33%, making our study practical for predicting seizures based on the pre-ictal state. With this approach, high-risk patients will be recognized, under control, and monitored sooner.