Automatic quiet sleep detection based on multifractality in preterm neonates: Effects of maturation

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Quiet sleep detection is crucial for monitoring preterm infant development.
• Traditional methods may not fully capture the complexity of neonatal electroencephalogram (EEG).
• Multifractal analysis offers a novel approach to characterize complex biological signals.

Purpose of the Study:

• To evaluate the efficacy of the multifractal formalism for detecting quiet sleep in preterm infants.
• To assess the influence of post-menstrual age on the performance of multifractal measures for sleep staging.
• To explore the relationship between EEG multifractality and neonatal maturation.

Main Methods:

• Utilized electroencephalogram (EEG) recordings from 25 healthy preterm infants.
• Applied multifractal analysis based on wavelet leaders to identify sleep epochs.
• Developed and evaluated classifiers to quantify performance, using Area Under the Curve (AUC).

Main Results:

• Multifractal analysis successfully identified quiet sleep epochs in neonatal EEG.
• Classifier performance was significantly influenced by infant post-menstrual age.
• Lowest AUC was observed in younger infants (≤ 31 weeks PMA), highest in full-term infants (≥ 37 weeks PMA).

Conclusions:

• Multifractal formalism is a viable tool for quiet sleep detection in preterm infants.
• Maturation of neonatal EEG leads to changes in multifractality, enhancing sleep stage distinguishability.
• Age-dependent performance highlights the importance of considering developmental stage in EEG analysis.