Epileptic seizure detection using heart rate variability from ambulatory ECG: a pseudoprospective study

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Seizure detection is crucial for epilepsy diagnosis and management.
• Current methods often rely on electroencephalography (EEG) in clinical settings.
• Ambulatory seizure detection using non-EEG signals is less explored.

Purpose of the Study:

• To investigate the feasibility of electrocardiography (ECG) for seizure detection in ambulatory settings.
• To develop a patient-independent, machine learning-based seizure detector using ambulatory ECG data.

Main Methods:

• Developed a machine learning seizure detector utilizing long-term ambulatory ECG monitoring data.
• Trained the model on data from 47 patients and evaluated it pseudoprospectively on 18 patients.
• Employed a patient-independent approach for broader applicability.

Main Results:

• The seizure detector performed better than chance for 14 out of 18 patients in the test set.
• Achieved an average sensitivity of 72% and an average specificity of 68% across the test cohort.
• Demonstrated improved performance in patients with focal epilepsy and those with significant heart rate changes during seizures.

Conclusions:

• Developed a novel, patient-independent seizure detection algorithm using ambulatory ECG data.
• Introduced a pseudoprospective evaluation framework for chronic ambulatory seizure monitoring.
• This approach holds potential for improving remote epilepsy monitoring and patient care.