413 - What’s wrong, baby? An acoustic analysis of cry triggers in newborns.

Poster Number: 413
Publication Number: 413.135

Adnan M. Al Ali, McGill University Faculty of Medicine and Health Sciences, Montreal, PQ, Canada; Charles Onu, McGill University, Montreal, PQ, Canada; Samantha Latremouille, Ubenwa Intelligence Solutions, Chambly, PQ, Canada; Wijdan Basfar, McGill University Faculty of Medicine and Health Sciences, Montreal, PQ, Canada; Eva B. Senechal, McGill University, Montréal, PQ, Canada; Doina Precup, McGill University & Mila, Montreal, PQ, Canada; Guilherme Sant'Anna, McGill University Health Center, MONTREAL, PQ, Canada

Background: Newborn cry is an involuntary reflex used to communicate needs, with typical triggers including temperature control and hunger. However, differentiation of triggers is subjective. Understanding newborn cry could help guide effective care from both parents and healthcare workers. Indeed, acoustic analyses of neonatal cry have demonstrated potential to classify a variety of health conditions or response to different stimuli.

Objective: To describe the acoustic characteristics of cries originating from various triggers in healthy term newborns.

Design/Methods: As part of a large prospective study of neonatal cry, at least one cry was recorded from healthy term newborns (36wks gestational age) for 30sec-3min using a Samsung A10s smartphone and an in-house developed study application during a neurological examination performed on the day of discharge from the nursery. Research staff annotated the primary reason for the crying. Importantly, cries were not actively induced by the research staff. Cry triggers included: a) spontaneous (i.e., undressing/exposure to cold), b) discomfort (i.e., soiled diaper), c) hunger, or d) startled during the Moro reflex test. Recordings were first passed through an AI-based algorithm to remove noise and extract only cry activity, which was then segmented into single cry units (CU; one voiced expiration), from which acoustic features were extracted using an automated algorithm to obtain pitch (median, min, and max), CU duration, and pause duration between CUs. The acoustic features were compared between cry triggers using the Kruskal-Wallis test.

Results: A total of105 healthy term newborns (Table 1) were recruited from the nursery, from which 121 recordings were analyzed (5 lost due to technical issues). Recordings had median 28 [20–41 IQR] CUs extracted. The number of annotated CUs per type of trigger were: a) spontaneous, n=1779; discomfort, n=831; hunger, n=518; or startle n=746. Despite a considerable amount of overlap, all 5 acoustic features were significantly different between cry triggers (Table 2). Moreover, consistent differences were noted between cries triggered by discomfort (lower pitch and longer durations) and hunger (higher pitch and shorter durations; Figure 1).Conclusion(s): Neonatal acoustic cry analysis is a potentially useful tool for parents or healthcare workers to differentiate between triggers. This preliminary data will help in the ongoing study assessing cry features in different pathological conditions.
Adnan Al AliAdnan Al Ali -merged.pdf