A multi-feature method for real-time seizure detection in pediatric intensive care unit

Area of Science:

• Pediatric Neurology
• Artificial Intelligence in Medicine
• Critical Care Medicine

Background:

• Continuous electroencephalogram (cEEG) monitoring is crucial for detecting electroclinical seizures in pediatric intensive care units (PICUs).
• Timely and accurate seizure detection in PICU patients is challenging but vital for effective management.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) method for real-time automatic detection of electroclinical seizures in the PICU.
• To assess the sensitivity and specificity of the AI detection system.

Main Methods:

• An AI method was designed to analyze electroencephalogram (EEG), electromyography (EMG), and electrocardiography (ECG) signals.
• The system extracted multi-dimensional features, identified candidate signal fragments in real-time, and analyzed indicators for electroclinical seizures.
• The AI system was tested on 28 PICU patients with seizure data collected via cEEG monitoring.

Main Results:

• The AI algorithm analyzed 218.73 hours of EEG data from 28 PICU patients.
• The system achieved an overall detection sensitivity of 94% with a false detection rate of 0.18/h.
• No significant difference in performance was observed between focal and generalized seizures.

Conclusions:

• The developed AI detection system demonstrates high sensitivity and specificity for electroclinical seizures in the PICU.
• This AI method holds significant potential for real-time automatic seizure detection, enhancing care for critically ill children.
• The system reliably identified seizures by integrating brain activity, muscle movement, and heart rate signals.