335: Discriminating Postural Impairments with Non-Linear Metrics from Smartphones for low back pain.

Topic: Clinical Science
Subtopic: Pain category (ICD-11), Chronic primary musculoskeletal pain, Physiotherapy (e.g. exercise, rehab), Artificial Intelligence/Machine Learning

Background and Aims: Low back pain individuals are shown to have altered postural control compared to healthy individuals, yet traditional kinematic analyses have shown limited habiliy to distinguish between healthy and LBP populations. The complexity of postural sway can be quantified using nonlinear measures. Non-linear measures can discriminate healthy controls from pathological people (eg, multiple sclerosis) at an early stage, although linear measures cannot. Mobile phone sensors have demonstrated reliability comparable to laboratory-based motion capture systems. By combining mobile sensor data and non-linear movement analysis we aim to develop an accessible, low-cost tool for early identification of LBP risk.

Methods: 43 healthy young adults (age: 21.1 ± 3.36) and 4 adults with low back pain completed four static balance tasks in a random order. The data were collected simultaneously on the smartphone. Resultant acceleration for the sagittal and frontal plane were calculated using raw vertical/anteroposterior and vertical/mediolateral acceleration data, respectively. A multiscale sample entropy analysis was performed on each postural acceleration time series using 15-time scales. The area under the complexity curve was calculated to obtain the complexity index. Various machine learning algorithms will be evaluated for classification accuracy and validated through ROC analysis with 3D acceleration.

Results: The mixed-effects model revealed significant effects of Group, Plane, and their interaction on entropy (MSE), indicating that postural sway patterns differ across these conditions.
Post-hoc Tukey HSD tests identified that the healthy group exhibited significantly lower entropy values compared to individuals with low back pain, in the sagittal and frontal planes. Reinforcing the model’s ability to detect nuanced movement impairments.
These differences suggest that mobile sensor data can capture subtle neuromuscular control variations, even in asymptomatic individuals.

Conclusions: Smartphones may provide easy access to sensitive nonlinear measures of postural sway during challenging balance conditions in non-laboratory settings instead of 3D MoCap. These findings support the hypothesis and demonstrate the potential of mobile sensor-based entropy measures for early identification of LBP risk
The entropy metrics derived from mobile phone sensors are sensitive enough to distinguish between movement profiles, laying the foundation for AI-based screening tools that could identify individuals at risk of low back pain before symptoms appear.
Further research is required in various age groups and clinical populations

References: 1. Lipsitz, L. A. and Goldberger, A. L. JAMA, 1992; 267 (13): 1806-1809. 2. , M. A. and van Emmerik, R. E. A. J Sport Health Sci, 2016; 5 (1): 44-51. 3. German, D. and Bahat, H. S. J Manipulative Physiol Ther, 2021; 44 (9): 707-717. 4. Alberts, J. L., Thota, A., Hirsch, J., Ozinga, S., Dey, T., Schindler, D. D., Koop, M. M., Burke, D. and Linder, S. M. Med Sci Sports Exerc, 2015; 47 (10): 2233-2240. 5. J.J. Craig, A.P. Bruetsch, S.G. Lynch, J.M. Huisinga, Clin Biomech 49 (2017) 16-21. 6. M.A. Busa, S.L. Jones, J. Hamill, R.E. van Emmerik, Gait Posture 45 (2016) 7-11

Presenting Author: Edith Elgueta Cancino
Submitting Author: Edith Elgueta-Cancino

Senior Author: Luis Cofre-Lizama

IASP 2026 Global Year theme of Neuropathic Pain: No

• Luis Cofre-Lizama: PhD - Department of Nursing and Allied Health, School of Health, Swinburne University of Technology, Melbourne, Australia