Eye movement detection using electrooculography and machine learning in cardiac arrest patients

Area of Science:

• Neurology
• Biomedical Engineering
• Critical Care Medicine

Background:

• Neuroprognostication of comatose cardiac arrest (CA) patients is challenging, necessitating novel biomarkers.
• Eye movement is a potential indicator of arousal recovery due to shared neuroanatomical pathways.
• Manual quantification of eye movements from electroencephalogram (EEG) with electrooculography (EOG) is labor-intensive.

Purpose of the Study:

• To develop and train a machine learning algorithm for automated eye movement detection using EOG data.
• To assess the feasibility of using EOG for continuous eye movement quantification in post-CA patients.
• To establish eye movement as a potential biomarker for neurological recovery in CA survivors.

Main Methods:

• A retrospective, single-center cohort study involving post-CA patients undergoing EEG/EOG monitoring.
• Training a machine learning algorithm on 1-hour EOG data from 48 patients (145,800 samples).
• Evaluating algorithm performance on a separate test set of 12-hour EOG data from 24 patients using AUC, sensitivity, and specificity.

Main Results:

• The algorithm achieved high performance in detecting eye movements.
• Sensitivity was 94.0%, specificity was 82.0%, and Area Under the Curve (AUC) was 94.2%.
• The study included 72 eligible patients, with survival and command-following rates reported for training and test groups.

Conclusions:

• Automated eye movement detection from EOG is highly sensitive and specific in cardiac arrest patients.
• This technology can potentially automate the quantification of eye movements.
• Further research can explore the association between quantified eye movements and patient recovery outcomes.