Automated State Analyses: Proposed Applications to Neonatal Neurointensive Care

Abstract

The two principal challenges of neonatal physiologic monitoring device are: (1) the development of computational strategies that consider the rudimentary forms of neonatal sleep state especially for preterm infants and (2) any physiologic monitoring device for clinical applications in a neonatal intensive care setting must be small, portable, and user-friendly. Our multicenter neonatal sleep consortium has acquired more than 1,100 multihour EEG-sleep recordings on over 370 neonates, ranging from 24 to 44 weeks gestation. Each recording was visually-scored for state, arousals, movements, and rapid eye movements, which were used as templates when applying spectral analyses. The authors have defined a brain dysmaturity index to represent functional brain reorganization as the prenate matures to a full-term age; delayed or accelerated physiologic behaviors have been described for the preterm cohort when compared to the full-term group at the same postmenstrual age. Seven EEG-sleep measures comprise this index: quiet sleep percentage, sleep cycle length, rapid eye movements, arousals, spectral beta EEG energies, spectral EEG correlations, and a spectral measure of respiratory regularity. Linear and nonlinear computational algorithms are being developed to automate the computation of the dysmaturity index and to identify new feature types that correlate with dysmaturity. Automated neonatal sleep monitoring system can potentially improve neonatal neurointensive care by facilitating analyses of pervasive neonatal brain disorders expressed primarily as altered sleep state organization, and help predict altered developmental trajectories of children at higher risk for neurologic sequelae.