Fast Iterative Filtering based approach for Seizure Detection in EEG Signals

The increasing necessity for precise epilepsy diagnosis has amplified the demand for robust classification of EEG signals into seizure and non-seizure categories. In this study, we present a novel method employing the Fast Iterative Filtering (FIF) algorithm to tackle this critical challenge. FIF offers a computationally efficient approach by decomposing the EEG signals into Intrinsic Mode Functions (IMF’s). These IMF’s capture the inherent time-frequency characteristics of brain activity. Subsequently, informative statistical features are extracted from the decomposed IMF’s. These features effectively represent distinct characteristics of seizures and normal brain activity. By leveraging the feature extraction capabilities of FIF and the classification power of XGBoost, we aimed to achieve high-resolution and accurate seizure detection. It is shown that our proposed method can achieve a categorization accuracy of up to 95% for EEG datasets. This combination offers the potential of a highly accurate and practical tool for automated EEG seizure detection in clinical settings.