The use of high frequency oscillations (HFO) as biomarkers in epileptogenic zone

: Epilepsy is a neurological disorder characterized by seizures that present themselves in a recurring pattern. The disease arises from abnormal activity in certain neurons in the brain. Epilepsy can be observed at any age, whether in childhood or adulthood, and presents with varying symptoms in different individuals. Seizures in epilepsy are commonly observed as muscle spasms, loss of consciousness, and fainting. Consequently, these can impact people's daily lives and their ability to engage in activities. Epilepsy can persist throughout life; however, with proper treatment and medications, seizures can be controlled, allowing individuals to lead a normal life. In some cases where suitable, surgery may be considered for the patient. Epilepsy surgery is a serious procedure, and therefore, patients are carefully evaluated. A multidisciplinary team comprising neurologists and neurosurgeons collaborates to assess the patient's suitability for the surgical procedure. In epilepsy surgery, different procedures are applied based on the location and condition of seizure foci with the aim of reducing seizures. This study investigated the presence of high-frequency oscillations that could potentially serve as a precursor to seizures. The Electroencephalography (EEG) data has been categorized into three distinct classes: pre-seizure, moment of seizure, and post-seizure. Features were extracted using signal processing techniques, and classification was performed based on these derived statistical and structural features. Additionally, 14 channels deemed important for each class were identified, and a classification study was conducted based on these channels. Following the determination of High-Frequency Oscillation (HFO) presence, a classification process was carried out based on HFO-based features. For classification purposes, Bayesian, k-Nearest Neighbors, Support Vector Machines, Multilayer Perceptron, and Random Forest Classifiers were utilized. In the classification results, the Random Forest Classifier achieved an accuracy rate of 83% for statistical and structural features, 77.43% for HFO-based features, and 77.97% for channels.