The performance evaluation of the state-of-the-art EEG-based seizure prediction models

Abstract

The recurrent and unpredictable nature of seizures can lead to unintentional injuries and even death. The rapid development of electroencephalogram (EEG) and Artificial Intelligence (AI) technologies has made it possible to predict seizures in real-time through brain-machine interfaces (BCI), allowing advanced intervention. To date, there is still much room for improvement in predictive seizure models constructed by EEG using machine learning (ML) and deep learning (DL). But, the most critical issue is how to improve the performance and generalization of the model, which involves some confusing conceptual and methodological issues. This review focuses on analyzing several factors affecting the performance of seizure prediction models, focusing on the aspects of post-processing, seizure occurrence period (SOP), seizure prediction horizon (SPH), and algorithms. Furthermore, this study presents some new directions and suggestions for building high-performance prediction models in the future. We aimed to clarify the concept for future research in related fields and improve the performance of prediction models to provide a theoretical basis for future applications of wearable seizure detection devices.

The present and future of seizure detection, prediction, and forecasting with machine learning, including the future impact on clinical trials

“…In addition to these two more established technologies, there are numerous technologies in development that address the challenges of the established technology ( Table 1 ) ( 38 , 74 , 75 ). The placement of the electrodes for the Responsive Neurostimulation System requires localization of the seizures, which can be challenging in some cases ( 73 ).…”

Long-term reliable neural decoding based on flexible implantable microelectronics and machine learning for seizure prediction application

“…We quantitatively compared the performance of two groups by calculating two clinically relevant metrics, namely the false alarm rate (the average number of false alarms triggered per hour during the interictal period) and the early warning time (the average alarm duration before seizure onset), which were introduced in previous studies 71,[73][74][75] . The PEDOT group demonstrated an early warning time of 8 ± 2 minutes before seizure onset, which is earlier than that of the tungsten group (5 ± 1 minutes), as shown in Fig.…”

The potential of the Lempel–Ziv complexity of the EEG in diagnosing cognitive impairment in patients with temporal lobe epilepsy

“…Quantitative electroencephalogram refers to the extraction of meaningful features from the EEG signals to delineate local and/or global changes in brain dynamics. [7][8][9] It is believed that each person exhibits a unique signal pattern on EEG, like a fingerprint of the brain. Besides, the EEG network is relatively stable for each adult individual and may correspond to the genetic makeup, disease-associated functional alterations, and treatment outcomes.…”

“…Quantitative electroencephalogram refers to the extraction of meaningful features from the EEG signals to delineate local and/or global changes in brain dynamics 7–9 . It is believed that each person exhibits a unique signal pattern on EEG, like a fingerprint of the brain.…”