Audio- and video-based estimation of the sleep stages of newborns in Neonatal Intensive Care Unit

Abstract

Premature babies have several immature functions and begin their life under high medical supervision. Since the sleep organization diers across postmenstrual age, its analysis may give a good indication of the degree of brain maturation. However, sleep analysis (polysomnography or behavioral observation) is dicult to install, time consuming and cannot systematically be used. In this context, development of new ways to automatically monitor the neonates, using contactless modalities, is necessary. Therefore, this study presents an innovative non-invasive approach to semi-automatize the classication of infant behavioral sleep states. Methods First, three descriptors were extracted from audio and video recordings: vocalizations, motion and eye state of the baby. For this purpose, an original semi-automatic algorithm for the estimation of the eye state was proposed. Secondly, the three descriptors were used in order to obtain an estimation of the behavioral sleep states. Five classiers (K-Nearest Neighbors, Linear Discriminant Analysis, Support Vector Machine, Random Forest and Multi-Layer Perceptron) were compared to an expert annotation. Results Firstly, the comparison of the semi-automatic eye state estimation to manual annotations of 10 videos led to a mean accuracy of 99.4%. Secondly, sleep stage classication was performed. Best results were obtained with Random Forest, for Quiet Alert and Active Alert stages, with 93.5% and 99.0% of accuracy respectively. Conclusion The proposed method provides a high capacity to identify alert sleep stages but the dierentiation between Quiet Sleep and Active Sleep only by behavioral observations still remains a dicult task to achieve. Signicance Results presented in this paper are new since no similar approach was proposed in the literature in the context of neonatal intensive care unit. They augur well for the automatic sleep organization assessment to improve newborn care.

Estimating Sleep Stages Using a Head Acceleration Sensor

“…From these, we can suppose that our data acquired through this experiment reflect general data for healthy subjects. Excluding the third night for subject 3, the number of datasets was 15, but in comparison with the studies of Cabon et al [ 27 ] and Ran et al [ 28 ], this is still sufficient to create an automatic sleep classifier under age- and gender-restricted conditions. However, to create an automatic sleep classifier that can handle various patterns, such as other genders, ages, and people suffering from sleep disorders, the number of subjects would need to be increased.…”

Quiet Sleep Organization of Very Preterm Infants Is Correlated With Postnatal Maturation

“…This increase was significant (w-test, p < 0.01) for the whole preterm population studied. Median time spent in QS increased from 13.0% [IQR: [13][14][15][16][17][18][19][20]] to 28.8% [IQR: [27][28][29][30] in Group 1 (w test, p < 0.05) and from 17.0% ADI spent in QS, expressed in seconds, was lower at T1 in group 1 than in group 2 (u-test, p < 0.05) and increased between T1 and T2 for all preterm newborns except one. This increase was significant (p < 0.01) for the whole preterm population studied.…”

“…The other alternative includes detailed clinical observation as in neonatal intervention programs, the use of actigraphy or amplitude-integrated EEG and the recent introduction of new methods for automatic classification of neonatal sleep states based on EEG (12)(13)(14). Some interventions have been proposed for sleep promotion but their evaluation is limited by the absence of reliable longitudinal monitoring (15)(16)(17).…”

Preterm Newborn Presence Detection in Incubator and Open Bed Using Deep Transfer Learning

“…Moreover, video modality can be used for high-level analysis and proved to be relevant in several clinical applications in pediatrics, particularly with motion analysis [10]. Regarding clinical applications involving preterm newborns, video-based motion analysis has been investigated for instance for early cerebral palsy detection [11], estimation of sleep stages [12], or maturation characterization [13]. The latter is motivated by the fact that the motor activity evolves along with the age of the newborns [14].…”