Virginia Research Day 2025

Graduate Student Research Clinical

01 The Validity of a Pose Estimation App Compared to a Universal Goniometer to Measure Joint Position

Niels Gainsback; Ben Harris; Haley Highsmith; Sarah Howell; Kevin Chui; Kevin Parcetich Corresponding author: kparcetichii@radford.edu

Radford University

shoulder ROM. Eleven participants were trained to use the PET application on a mobile device and were instructed to capture joint position after researchers set a joint position using the universal goniometer. The joint position was recorded from the PET mobile application and compared to the set measurement obtained by the goniometer. Accuracy, inter-rater, and intra-rater reliability were determined. Results: Intraclass Correlation Coefficients (ICC) and Bland Altman analyses were performed to determine inter-rater reliability, intra-rater reliability, and observe trends between each of the 5 motions. Small differences were observed using the Bland Altman plots with an overall trend of overestimating joint position values using PET. Model 1 intra-rater reliability for abduction mid-range ICC of 0.597 (CI 0.083-0.869, p=0.014), IR end range ICC of 0.542 (CI -0.011-0.849, p=0.013), and ER beginning range ICC of 0.525 (CI –0.059-0.844, p=0.041). Model 2 intra rater reliability for abduction mid-range ICC of 0.548 (CI –0.044-0.855, p=0.036) and ER end-rage ICC of 0.617 (CI 0.073-0.880, p=0.018). Inter-rater reliability

between “raters”, findings report flexion beginning range ICC of 0.505 (CI 0.121-0.999, p<0.001), flexion mid-range ICC of 0.361 (CI 0.054-0.998, p=.004), flexion end-range ICC of 0.574 (CI 0.132 0.999, p=0.004), and abduction beginning-range ICC of 0.574 (CI 0.155-0.999, p<0.001). Implications: The initial pilot data suggests pose estimation technology may overestimate shoulder joint positions and can provide reliable measurements for select positions with large confidence intervals. Limitations of this study include assess over larger sample size with varying demographics and anthropometric characteristics, use in different joints, and in those with pathologies that impair ROM. Approved by Radford University IRB, protocol #2024 026-RUC. Authors received patient consent to use data for report.

Introduction: Joint position is measured with a universal goniometer to quantify impaired movement and monitor the progress of physical rehabilitation treatments. Pose Estimation Technology (PET) has become a more common method of capturing joint position, using computer algorithms and deep learning models to recognize and track anatomical landmarks. PET has implications in the field of telehealth, allowing for increased patient access, however, accurate and reliable virtual physical assessments are required to replace in-person assessments. Purpose: The purpose of this study is to determine the accuracy and reliability of a mobile PET application when measuring different joint positions of the glenohumeral joint. Methods: The right shoulder of two asymptomatic “models” was used to capture joint positions across beginning, middle, and end ranges of motion for 5 different shoulder motions (flexion, extension, abduction, internal and external rotation). Two researchers, proficient in the use of a universal goniometer, set shoulder joint positions at different points in the available

Table of Contents

18