(F-218) Characterizing Infant Rolling Movements: Comparing Laboratory Techniques with At-Home Video

Introduction:: Rolling is an important infant milestone that encourages muscle development and postural control needed to achieve more advanced movements like sitting and crawling [1]. Assessing infant rolling patterns is crucial in the early identification of motor development delay [2] and presently only 3.1% of infants receive an early diagnosis [3]. Previously, six unique movements infants utilize to achieve a roll have been defined (Fig.1A) [4], but identifying these movements is limited to a laboratory setting where a costly motion capture system and advanced biomechanics knowledge is needed. While these techniques are considered the gold-standard due to the high levels of precision [5], we must develop methodology that can more easily be used by clinicians to assist in the diagnosis of possible motor delays. Therefore, the purpose of this study is to develop video-based methods in which the six movements of infant rolling defined through motion capture can be identified through video alone.


Materials and Methods:: Marker-based motion capture data (Vicon, 100Hz) and GoPro video was collected simultaneously from 14 healthy infants (6.6±0.7 mos; 6F). Infants were placed supine on a flat firm playmat for five minutes or until they achieved a complete supine to prone roll (Fig.1D). Custom MATLAB code was created to analyze the motion capture data for 24 rolling movements, defined as the supine to lateral rotation of the torso [4]. The speed of each limb and the trunk was calculated and normalized to the speed of the torso for the initial 25% of the movement. Time series plots determined the coordination of the limbs and defined the rolls as one of the six movements. Video-based methodology was developed to identify the six coordinated movements from video alone. Three researchers and one clinician reviewed the corresponding 24 videos and a Fleiss’ Kappa statistical analysis was performed to determine the accuracy of our video-based methodology compared to the gold-standard of motion capture.

Results, Conclusions, and Discussions:: The overall reliability score of the Fleiss’ Kappa was 0.764, making the agreement substantial. This indicates that our video identification methods match the motion capture gold-standard with significant agreement. This study is the first that we know of that distinguishes coordinated movement patterns of infant rolling through video alone. While other studies have demonstrated rolling categorization with advanced laboratory equipment [4], these methods can be hard to accomplish quickly and accurately when looking at a long-term study of infant rolling development or for clinicians diagnosing motor delays. The overall impact of this work is that clinicians can now have another resource in identifying infant rolling patterns and techniques, and thus assist in the diagnosis of possible motor delays and disorders. Researchers can also implement these techniques to increase the feasibility of studying infant development, specifically rolling, more thoroughly.

Acknowledgements (Optional): : The data used in this study was collected as part of a research contract with Iron Mountains LLC. They had no role in the design, data collection, analysis, interpretation, or drafting of the abstract.

References (Optional): : [1] Altmann TR & Hill DL. American Academy of Pediatrics, 2019. [2] Harris SR & Heriza CB. Physical Therapy, 1987. [3] Early Childhood Technical Assistance Center, 2022. [4] Kobayashi Y, at al. Experimental Brain Research, 2016. [5] Colyer SL, et al. Sports Medicine, 2018.