Enhancing Epileptic Seizure Detection Accuracy Using YOLOv11 Classification

Epilepsy is a long-term neurological condition that causes repeated seizures, demands precise and timely detection to facilitate effective treatment and management. This study introduces an innovative approach for automated seizure detection using the CHB-MIT Scalp EEG Database, a widely utilized resource of pediatric EEG recordings The CHB-MIT Scalp EEG Database is a widely recognized dataset used for seizure detection and classification, comprising EEG recordings from 23 pediatric patients aged 1.5 to 22 years. Collected using the International 10-20 electrode system at a 256 Hz sampling rate, the dataset includes 182 annotated seizures stored in EDF format. To enhance its usability for deep learning applications, we developed a preprocessing pipeline that converts raw EEG signals into image representations, enabling the use of computer vision-based models such as YOLOv11. The dataset, consisting of 6,579 labeled images (seizure and non-seizure), was augmented using techniques like brightness adjustment, grayscale conversion, and noise injection. The images were split into training (92%), validation (4%), and test (4%) subsets. Our YOLOv11-based model achieved an accuracy of 98.8%, precision of 98.7%, recall of 98.8%, and an F1-score of 98.7%, demonstrating its effectiveness in seizure classification. These results highlight how deep learning can help automate seizure detection, making early diagnosis and treatment for epilepsy patients more effective.