Adaptive threshold algorithm for detecting EEG-interburst intervals in extremely preterm neonates

Area of Science:

• Neonatal neurology
• Medical signal processing

Background:

• Electroencephalograms (EEG) are crucial for monitoring preterm infants' neurological status.
• Accurate burst detection in neonatal EEG is essential for diagnosing and managing conditions like hypoxic-ischemic encephalopathy.
• Current manual EEG analysis is time-consuming and subject to inter-rater variability.

Purpose of the Study:

• To develop and evaluate an adaptive threshold algorithm for automated burst detection in neonatal EEG.
• To assess the algorithm's performance on unselected, real-world clinical EEG data from preterm infants.
• To compare the algorithm's accuracy against expert human raters.

Main Methods:

• An adaptive threshold algorithm was developed for burst detection in EEG signals.
• The algorithm was tested on a dataset of 30 clinical EEG recordings from preterm infants, without preselection for quality.
• Performance metrics included inter-rater agreement (kappa), sensitivity, and specificity, compared to a clinical expert.

Main Results:

• The algorithm achieved substantial inter-rater agreement (kappa = 0.73).
• Performance against a clinical expert showed similar agreement (kappa = 0.73), with high sensitivity (0.90) and specificity (0.95).
• The algorithm demonstrated robust performance on unselected, real-world clinical data.

Conclusions:

• The adaptive threshold algorithm is a practical and effective tool for automated burst detection in neonatal EEG.
• The algorithm's performance is comparable to that of experienced human raters.
• This automated approach can aid in the efficient and accurate assessment of preterm infants' neurological function via EEG.