A Simple Back Shape Model for Back Curvature Evaluation at the Sagittal Plane Using a Single 2D Image

Low back pain is a prevalent health problem. To provide appropriate treatment for it, posture evaluation including flexibility of the spine and hip joints is important. However, this evaluation can be subject to variability between therapists. Therefore, an easy method that help therapists carry out objective evaluations is required. We propose a simple back shape model for image-based evaluation of forward flexion and backward extension, which are postures crucial for spine and hip-joint assessment. In an experiment involving 94 healthy adults, the model accurately estimated the boundaries of the back segments with an error of 4 cm compared with those annotated by a physical therapist. Moreover, we confirmed that image features based on the model, such as lordosis and kyphosis angles and curvature of the back, contributed to accurate estimation of posture evaluation labels representing what therapists observe in assessing forward flexion and backward extension with 62.1-89.7% accuracy.Clinical Relevance-The proposed method offers a simple and radiation-free measurement for the spine and hip joints at forward flexion and backward extension postures by utilizing a single 2D RGB image, allowing patients to remain fully clothed. This will contribute to the promotion of objective assessments by physical therapists in clinical practice.