Dynamic assessment of spine movement patterns using an RGB-D camera and deep learning
Jessica Wenghofer1, Kristen He Beange2, Wantuir C Ramos1
1School of Human Kinetics, Faculty of Health Sciences, University of Ottawa, Ottawa, ON, Canada.
Journal of Biomechanics
|March 5, 2024
Summary
This study introduces a deep learning (DL) markerless motion capture system using RGB-D cameras for objective spine movement assessment in low back pain patients, improving diagnostic accuracy.
Area of Science:
- Biomedical Engineering
- Computer Vision
- Rehabilitation Medicine
Background:
- Functional limitations in low back pain are subjectively assessed, risking misdiagnosis and prolonged pain.
- Objective kinematic assessment is crucial for accurate diagnosis and treatment of spinal conditions.
Purpose of the Study:
- To develop and validate a deep learning (DL) markerless motion capture system for objective spine kinematics measurement.
- To assess spine flexion-extension (FE) movements in patients with low back pain.
Main Methods:
- A DL semantic segmentation algorithm was developed to segment the back and pelvis into anatomical classes.
- A kinematic framework utilized these segmentations to measure spine kinematics during FE.
- Data was collected using an RGB-D camera, with validation against an optical motion capture (OPT) system.
Main Results:
- The DL segmentation algorithm demonstrated high accuracy.
- Root mean square error (RMSE) for lumbar kinematics was < 4° compared to ground truth.
- RMSE between markerless and OPT kinematics was < 6°.
Conclusions:
- The proposed markerless motion capture system is feasible for assessing spine FE movement in clinical settings.
- This technology offers a potential solution for objective assessment of functional limitations in low back pain.
- Future research will expand movement analysis and demographic testing.
