Automated Posture Assessment for Non-specific Low Back Pain from 2D Images

Assessing the flexibility of spine and hip joints is essential for providing appropriate treatment for non-specific low back pain. In general, posture assessment is subjective, and its reliability depends on the experience of physiotherapists. Therefore, there is a need for an automated tool that enables physiotherapists with any level of experience to carry out standardized posture assessments. In this paper, we propose image processingbased automated assessment models of multi-segmental flexion, extension, and rotation postures to comprehensively assess the movements in sagittal and axial planes. In order to achieve detailed movement assessment in the sagittal plane, we used body trunk shape-based features in addition to key points-based features such as angles of joints. Using 94 healthy participants’ data annotated by a musculoskeletal rehabilitation expert, we evaluated the accuracy of our assessment models and confirmed improvements by use of our body trunk shape-based features. Additionally, we discuss the pilot deployment of our models as a supporting tool for posture assessment.