Recognizing and Predicting Neonatal Pain in Preterm Intensive Care Unit: a Study Protocol

A robust body of research reports that pain is an important determinant of neurodevelopmental outcome in sick preterm and term infants. The newborn patient is uniquely exposed to pain due to underdeveloped nociceptive modulation, high prevalence of painful procedures, and lack of effective methods to assess and predict pain intensity and duration. More than forty neonatal pain scales have been developed, which testifies to both the magnitude of the problem and the complexity of the task. Trained neural network models have been shown to effectively discriminate facial expressions and human body language. We propose the approach of combining the classification of facial and body images by neural networks with professional evaluation and objective data, e.g. patient data, pathology, and vital parameters. Such an approach would offer critical advantages over scaling tools, as images could be acquired sequentially, allowing continuous rather than intermittent monitoring, even in absence of a human supervisor, and enabling quantification of neonatal pain in the domains of time and intensity.