Reliability and Agreement of a Smartphone-Based Markerless Analysis System (Sports2D) for Joint Range of Motion Measurement
10.5763/kjsm.2025.43.4.254
- Author:
Liun KIM
1
;
HunMin KIM
;
Seungjun KO
;
Joo Hyun LEE
;
Jinkyung CHO
;
Eunwook CHANG
Author Information
1. Department of Kinesiology, Inha University, Incheon, Korea
- Publication Type:Clinical Article
- From:The Korean Journal of Sports Medicine
2025;43(4):254-264
- CountryRepublic of Korea
- Language:English
-
Abstract:
Purpose:The goniometer, the gold standard for range of motion (ROM) assessment, has significant limitations including physical contact requirements and high interrater variability. While markerless motion capture using deep learning (e.g., Sports2D) offers a noncontact alternative, its clinical reliability remains unestablished. This study evaluated Sports2D's reliability for ROM measurement using smartphone cameras compared to conventional goniometry.
Methods:Fifty healthy adults (25 males, 25 females; mean age, 22.42±2.58 years) participated. Ten joint ROM types (17 total measurements including left/right) were measured using both traditional goniometry (by an experienced clinician) and Sports2D (smartphone camera at 3 m distance). Reliability and agreement were analyzed using concordance correlation coefficient (CCC), Pearson correlation, and Bland-Altman plots.
Results:Sports2D demonstrated very high reliability for elbow flexion (CCC, 0.906−0.955), neck extension (CCC, 0.928), and right hip flexion (CCC, 0.920). Good reliability (CCC, ≥ 0.8) was observed for neck flexion, left knee flexion, and trunk flexion. However, all ankle measurements showed clinically unacceptable reliability (CCC, ＜ 0.3), including dorsiflexion (CCC, 0.072−0.099) and plantarflexion (CCC, 0.174–0.211). Bland-Altman plots confirmed proportional bias for ankle measurements, particularly at smaller angles.
Conclusion:Sports2D is a reliable noncontact tool for measuring ROM in specific joints including elbow, neck, knee, hip, and trunk. However, it is unreliable for ankle joint assessment due to significant measurement discrepancies. These findings highlight the need for selective clinical application of markerless motion capture based on joint-specific validation.
