A novel dictionary for neonatal EEG seizure detection using atomic decomposition.

Area of Science:

• Neonatal neurology
• Biomedical signal processing
• Machine learning in healthcare

Background:

• Automated seizure detection in neonatal electroencephalogram (EEG) is crucial for timely intervention.
• Existing methods face challenges in accurately identifying seizures in complex neonatal brain activity.

Purpose of the Study:

• To enhance a previously developed atomic decomposition method for EEG seizure detection.
• To introduce a novel time-frequency (TF) dictionary optimized for neonatal EEG seizure signals.

Main Methods:

• Development of a new TF dictionary tailored for neonatal EEG seizure characteristics.
• Comparison of the proposed dictionary against Gabor, Fourier, and wavelet dictionaries.
• Validation using real-world neonatal EEG data.

Main Results:

• Dictionary selection significantly impacts EEG seizure detection accuracy.
• The proposed TF dictionary demonstrated a 10% improvement in seizure detection accuracy.
• A 5% improvement in the area under the Receiver Operator Characteristic (ROC) curve was observed.

Conclusions:

• The developed TF dictionary offers superior performance for automated neonatal EEG seizure detection.
• This advancement holds potential for improved clinical diagnosis and patient outcomes in neonatology.