Enhancing the Assessment of Deglutition Function in Preterm Infants With Mechano-Acoustic Analysis and Machine Learning: A Narrative Review

Purpose: Clinical indicators of deglutition dysfunction within the context of maturation are not well defined for preterm infants. Mechano-acoustic analysis utilizing cervical auscultation or accelerometry has shown potential for an accurate classification of swallow physiology across the lifespan. Machine learning algorithms increase diagnostic performance and facilitate the analysis of more complex deglutition data in adult and pediatric populations. This narrative review will investigate the feasibility and usability of mechano-acoustic analysis combined with machine learning to identify indicators of deglutition impairment in preterm infants to increase the diagnostic accuracy and clinical prediction of clinical swallow evaluations. Method: Databases searched included PubMed and Ovid. No filters were placed on the year of publication. Results: Twelve relevant records were retrieved for this review article. Preliminary studies investigating maturational changes in the mechano-acoustic features of deglutition have yielded significant findings. Research utilizing machine learning to support mechano-acoustic analysis in preterm infants is lacking. There are no published studies investigating the indicators of deglutition dysfunction in preterm infants and no normative data for healthy, term, nondysphagic neonates. Conclusions: Mechano-acoustic analysis is a feasible technique to investigate deglutition performance in preterm infants. Identification of normative values, temporal correlation of signals with deglutition kinematics, and standardization of relevant features in future longitudinal studies will enhance clinical utility. Further study is needed to determine the diagnostic performance of machine learning algorithms to enhance the classification and prediction of deglutition function in preterm infants.