Neonatal apnea and hypopnea prediction in infants with Robin sequence with neural additive models for time series.

Neonatal apneas and hypopneas present a serious risk for healthy infant development. Treating these adverse events requires frequent manual stimulation by skilled personnel, which can lead to alarm fatigue. This study aims to develop and validate an interpretable model that can predict apneas and hypopneas. Automatically predicting these adverse events before they occur would enable the use of methods for automatic intervention. We propose a neural additive model to predict individual occurrences of neonatal apnea and hypopnea and apply it to a physiological dataset from infants with Robin sequence at risk of upper airway obstruction. The dataset will be made publicly available together with this study. Our proposed model allows the prediction of individual apneas and hypopneas, achieving an average AuROC of 0.80 when discriminating segments of polysomnography recordings starting 15 seconds before the onset of apneas and hypopneas from control segments. Its additive nature makes the model inherently interpretable, which allowed insights into how important a given signal modality is for prediction and which patterns in the signal are discriminative. For our problem of predicting apneas and hypopneas in infants with Robin sequence, prior irregularities in breathing-related modalities as well as decreases in SpO2 levels were especially discriminative. Our prediction model presents a step towards an automatic prediction of neonatal apneas and hypopneas in infants at risk for upper airway obstruction. Together with the publicly released dataset, it has the potential to facilitate the development and application of methods for automatic intervention in clinical practice.