A review of epilepsy detection and prediction methods based on EEG signal processing and deep learning

Frontiers in Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: Nov. 15, 2024

Epilepsy is a chronic neurological disorder that poses significant challenges to patients and their families. Effective detection prediction of epilepsy can facilitate patient recovery, reduce family burden, streamline healthcare processes. Therefore, it essential propose deep learning method for efficient epileptic electroencephalography (EEG) signals. This paper reviews several key aspects EEG signal processing, focusing on prediction. It covers publicly available datasets, preprocessing techniques, feature extraction methods, learning-based networks used in these tasks. The literature categorized based independence, distinguishing between patient-independent non-patient-independent studies. Additionally, the evaluation methods are classified into general classification indicators specific criteria, with findings organized according cycles reported various review reveals important insights. Despite availability public they often lack diversity types collected under controlled conditions may not reflect real-world scenarios. As result, tend be limited fully represent practical conditions. Feature network designs frequently emphasize fusion mechanisms, recent advances Convolutional Neural Networks (CNNs) Recurrent (RNNs) showing promising results, suggesting new models warrant further exploration. Studies using data generally produce better results than those relying data. Metrics typically perform though future research should focus latter more accurate evaluation. kept 1 h, most studies concentrating intervals 30 min or less.

Language: Английский

Epileptic Seizure Detection in Neonatal EEG Using a Multi-Band Graph Neural Network Model

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(21), P. 9712 - 9712

Published: Oct. 24, 2024

Neonatal seizures are the most common clinical presentation of neurological dysfunction, requiring immediate attention and treatment. Manual detection seizure events from continuous electroencephalogram (EEG) recordings is laborious time-consuming. In this study, a novel graph-based method for automated neonatal proposed. The proposed aims to improve performance by thorough representation multi-channel EEG signals adaptive classification multi-band graph representations. To achieve this, band-wise feature extraction performed on raw provide more detailed information classification. addition, model, namely neural network (MBGNN), proposed, which utilizes mechanism can take full advantage representations performance. evaluated using 39 neonates Helsinki database. MBGNN model gives an average area under receiver operating characteristic curve (AUC) 99.11%, positive predictive value (PPV) 95.34%, negative (NPV) 96.66%. experimental results show that could fully exploit facilitate seizure/non-seizure epochs, making it appealing patient-specific applications.

Language: Английский