Rationale:
People with epilepsy are at risk of seizure-induced and sleep-related respiratory abnormalities. However, many hospitals do not directly record respiration via chest belts during long-term monitoring (LTM) in epilepsy monitoring units (EMU). Electrocardiograms (ECG) are always recorded during LTM, but the EMU-ECG electrodes are not placed in standard locations. The goal of this study is to leverage the EMU-ECG to approximate respiratory function in retrospective EMU recordings, particularly during sleep and surrounding seizures.
Methods:
Using a wearable device, ECG and respiratory waveforms were recorded concurrently from healthy volunteers over multiple days. Known ECG-derived respiration (EDR) algorithms were applied to the ECGs, and each EDR waveform was compared to the thoracic respiratory belt waveform. Data was stratified by signal quality, heart rate, and ECG morphology. The best performing methods were tested on peri-ictal recordings from participants wearing the same device during their LTM admission.
Results:
EDR methods were run on 22,080 30-second segments of which 91.5% (20,192) were in a physiological heart rate range (40-180 bpm). The EDR methods exhibited an average respiratory waveform peak correlation of >0.85, and spectral coherence of the EDR and respiratory waveforms of >0.8. The EDR algorithms performed best on windows of 30-seconds, compared to 1-minute, 2-minutes, and 5-minutes.
Conclusions:
This project will provide the epilepsy field with a validated ECG-based tool to study respiratory function in EMU recordings, specifically during sleep and surrounding seizures. This will allow researchers to better understand the cardio-respiratory changes surrounding seizures that could ultimately lead to sudden unexpected death in epilepsy (SUDEP).
Funding: SUNY Upstate Hendricks Pilot Grant