Towards a wearable multi-modal seizure detection system in epilepsy: A pilot study

Area of Science:

• Neurology
• Biomedical Engineering
• Medical Devices

Background:

• Epilepsy monitoring often relies on traditional electroencephalography (EEG).
• Wearable technology offers potential for continuous, multi-modal physiological data collection.
• Developing effective automated seizure detection strategies is crucial for patient care.

Purpose of the Study:

• To investigate the feasibility of wearable multi-modal monitoring in epilepsy.
• To identify effective strategies for automated seizure detection using combined data streams.
• To compare multi-modal monitoring with uni-modal EEG for seizure detection.

Main Methods:

• Thirty patients with suspected epilepsy underwent video-EEG monitoring with a wearable multi-modal setup.
• Continuous data recording included electroencephalography (EEG), electrocardiography (ECG), and accelerometry (ACM).
• A support vector machine (SVM) algorithm was trained for cross-modal seizure detection.

Main Results:

• SVM classification achieved 84% sensitivity for focal tonic seizures with a false alarm rate (FAR) of 8/24h in one patient.
• For focal non-motor seizures, sensitivities of 100% were achieved with FARs of 13/24h and 5/24h in two patients.
• Visual analysis of multi-modal data informed future seizure detection strategy development.

Conclusions:

• Wearable multi-modal monitoring is a feasible approach for epilepsy assessment.
• Automated seizure detection benefits significantly from integrating multiple data modalities (EEG, ECG, ACM) over uni-modal EEG.
• This unique combination of wearable sensors provides valuable insights for future epilepsy monitoring research.