Complexity analysis of neonatal EEG using multiscale entropy: applications in brain maturation and sleep stage classification

Automated analysis of the electroencephalographic (EEG) data for the brain monitoring of preterm infants has gained attention in the last decades. In this study, we analyze the complexity of neonatal EEG, quantified using multiscale entropy. The aim of the current work is to investigate how EEG complexity evolves during electrocortical maturation and whether complexity features can be used to classify sleep stages. First , we developed a regression model that estimates the postmenstrual age (PMA) using a combination of complexity features. Then, these features are used to build a sleep stage classifier. The analysis is performed on a database consisting of 97 EEG recordings from 26 prematurely born infants, recorded between 27 and 42 weeks PMA. The results of the regression analysis revealed a significant positive correlation between the EEG complexity and the infant’s age. Moreover, the PMA of the neonate could be estimated with a root mean squared error of 1.88 weeks. The sleep stage classifier was able to discriminate quiet sleep from nonquiet sleep with an area under the curve (AUC) of 90%. These results suggest that the complexity of the brain dynamics is a highly useful index for brain maturation quantification and neonatal sleep stage classification.