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

Area of Science:

• Biomedical Engineering
• Clinical Biomechanics
• Rehabilitation Science

Background:

• Low back pain is a widespread issue requiring accurate posture evaluation.
• Current posture assessments by therapists can lack objectivity and consistency.
• Objective evaluation of spine and hip joint flexibility is crucial for effective low back pain management.

Purpose of the Study:

• To develop a simple, image-based back shape model for objective posture evaluation.
• To assess the accuracy of the model in estimating back segment boundaries and posture.
• To determine if model-derived features correlate with therapist-observed posture labels.

Main Methods:

• A novel, simple back shape model was created for image-based analysis.
• Experiments were conducted on 94 healthy adults using the model.
• The model's accuracy was compared against physical therapist annotations for back segment boundaries.
• Image features like lordosis, kyphosis, and curvature were extracted.

Main Results:

• The back shape model estimated back segment boundaries with a 4 cm error compared to therapist annotations.
• Model-derived features accurately estimated posture evaluation labels (62.1-89.7% accuracy).
• The method provides radiation-free, clothed assessments using a single 2D RGB image.

Conclusions:

• The proposed model offers a simple, objective, and radiation-free tool for assessing spine and hip joint posture.
• This technology can enhance the objectivity of physical therapist assessments in clinical practice.
• It supports improved treatment for low back pain by providing reliable posture data.