Non-EEG-Based Seizure Detection: A Comprehensive Review

This paper presents a review of non-electroencephalogram (EEG) modalities for seizure detection. Seizure detection is crucial for individuals with epilepsy, however traditional EEG-based methods pose limitations in terms of accessibility and usability. This study explores alternative signals, such as heart rate variability (HRV), electrocardiogram (ECG), accelerometry, and others, for robust and non-invasive seizure detection. The Paper focuses on various non-EEG-based modalities for seizure detection, including wearable devices, biosensors, smartphone applications, and machine learning algorithms. Each modality is evaluated based on its potential advantages, such as portability, continuous monitoring capabilities, and real-time alerts. This work also investigates recent advancements in non-EEG-based seizure detection, including the integration of multiple physiological signals, such as heart rate variability, skin conductance, and motion on patterns, to improve detection accuracy and reliability. Previously reported results are compared based on the technologies utilized, including wearable multimodal convulsive seizure detectors incorporating electrodermal activity, accelerometry biosensors, and heart rate variability measured using ECG. These results are evaluated on different performance metrics like sensitivity, false alarm rate (FAR), specificity and accuracy. It is found that automated HRV algorithm with matching pursuit (MP) and wigner–ville distribution (WVD) algorithm, has 100% sensitivity and 99.91% Specificity for detecting seizures, while FAR was low 0.2 events/day for wearable multimodal convulsive seizure detectors.