Videos of common activities reveal movement patterns associated with biopsychosocial low back function in a worldwide...

Low back pain (LBP) is frequently classified as non-specific, which reduces the capacity of the healthcare system to offer targeted rehabilitation. Previous work has suggested a link between motor control and low back dysfunction. Identifying motor control phenotypes indicative of dysfunction often requires complex and costly laboratory equipment. Recent advancements in computer vision and the widespread availability of smartphones have made human motion capture more accessible. This study aimed to examine whether outcomes derived from consumer-grade video and open-source pose estimation tools are associated with motor control patterns linked to self-reported low back dysfunction.

A self-guided online questionnaire was employed to gather data from 448 participants, worldwide. Participants completed validated questionnaires and video-recorded themselves performing four functional movements. Pose-derived kinematic features were extracted and reduced using principal component analysis (PCA), followed by a stepwise linear regression modelling to examine associations between movement features and individual participant reported outcome measures as well as a composite index of low back function. Participants were split into low and high function groups.

PCA-derived features of movement were significantly associated with measures of disability, kinesiophobia, pain catastrophizing, and physical activity. Trunk flexion demonstrated the strongest association with the composite index (R² = 0.71). The results between low vs high function participants depict biomechanically relevant differences (i.e. reduced movement speed and range of motion) typically found in the low back pain (LBP) population.

Results highlight the feasibility of using consumer-grade video and open-source pose estimation tools for large-scale biomechanical data collections, to enhance our understanding of LBP. Although strong associations were observed between video-derived movement features and self-reported dysfunction, prospective validation and external testing may be required before clinical screening performance can be established. With appropriate validation, this approach has the potential to support the development of a scalable and accessible digital movement assessment tool for low back dysfunction.