A Multi-Sensor, Multi-Movement Exploratory Study of Motion Tape Strain Data for Low Back Pain Classification

Objective assessment of low back pain (LBP) is challenging due to subtle, task-dependent movement impairments that are poorly captured by existing sensing technologies. Motion Tape (MT), which is a self-adhesive elastic fabric skin strain sensor, enables skin-conforming measurement of localized biomechanical strain during functional movement, but its discriminative utility for LBP remains unclear. We examine this question in a multi-sensor, multi-movement setting and analyze whether MT signals encode discriminative structure that distinguishes individuals with LBP from healthy controls. Using data from 20 participants performing 19 functional movements with six sensors, we evaluate movement-specific classification under a leave-pair-out protocol and examine which movements, sensor placements, and features are most informative. Our analysis reveals that group separation is highly selective: only a small subset of movements, most notably forward flexion, consistently supports accurate classification, while many movements remain at near-chance level. We find that temporal dynamics features help in resolving difficult cases that global strain statistics fail to separate, and that informative signals are spatially localized to the lower lumbar spine. In contrast, pretrained time-series foundation models show negligible sensitivity to participant-level structure in MT signals. Overall, the findings from this exploratory study establish when and how MT sensing can effectively differentiate individuals with LBP from healthy controls, providing a principled foundation for larger-scale validation.