Introduction to neonatal facial pain detection using common and advanced face classification techniques

[abstract]

Assessing pain in neonates is a challenging problem. Neonates cannot describe their pain experiences but must rely exclusively on the judgments of others. Studies demonstrate, however, that proper diagnosis of pain is impeded by observer bias. It has therefore been recommended that neonatal pain assessment instruments include evaluations that have bypassed an observer. In this chapter, we describe the Infant COPE (Classification of Pain Expressions) project and our work using face classification to detect pain in a neonate's facial displays. We begin by providing an introduction to face classification that includes an outline of some common and advanced algorithms. We then describe a small database we designed specifically to investigate classifier performance in this problem domain. This is followed by a summary of the experiments we have performed to date, including some preliminary results of current work. We believe these results indicate that the application of face classification to the problem of neonatal pain assessment is a promising area of investigation.

[full paper]