| تعداد نشریات | 21 |
| تعداد شمارهها | 1,273 |
| تعداد مقالات | 11,617 |
| تعداد مشاهده مقاله | 82,176,922 |
| تعداد دریافت فایل اصل مقاله | 121,272,739 |
Validation of AI-Based Finger Joint Angle Estimation Using MediaPipe and OpenPose during Computer Mouse Use | ||
| Journal of Biomedical Physics and Engineering | ||
| مقالات آماده انتشار، اصلاح شده برای چاپ، انتشار آنلاین از تاریخ 10 خرداد 1405 | ||
| نوع مقاله: Original Research | ||
| نویسندگان | ||
| Ali Sahraneshin Samani1، 2؛ Alireza Choobineh* 3؛ Mohammad Abdoli-Eramaki4؛ Mohsen Razeghi5؛ Fardin Negahdari1، 2 | ||
| 1Student Research Committee, Shiraz University of Medical Sciences, Shiraz, Iran | ||
| 2Department of Ergonomics, School of Health, Shiraz University of Medical Sciences, Shiraz, Iran | ||
| 3Research Center for Health Sciences, Institute of Health, Shiraz University of Medical Sciences, Shiraz, Iran | ||
| 4School of Occupational and Public Health, Toronto Metropolitan University, Toronto, Canada | ||
| 5Department of Physiotherapy, Shiraz University of Medical Sciences, Shiraz, Iran | ||
| چکیده | ||
| Background: Accurate quantification of finger joint kinematics is essential for ergonomic assessment, particularly in computer-based work environments where repetitive hand movements may contribute to musculoskeletal disorders. Objective: This pilot study evaluates the performance of two Artificial Intelligence (AI) based markerless pose estimation systems; MediaPipe and OpenPose for measuring the Proximal Interphalangeal (PIP) joint angle of the index finger during computer mouse use, using an electrogoniometer as the ground truth. Material and Methods: In this methodological validation study, thirteen right-handed participants performed a 3-minute unscripted computer mouse task. At the same time, finger joint angles were simultaneously recorded via an electrogoniometer and two commercial digital cameras (frontal and lateral views). Joint angles were extracted using Python-based implementations of MediaPipe and OpenPose. Results: Results showed that MediaPipe with lateral view achieved the highest validity, with an RMSE (Root Mean Square Error), of 4.22°, and 92.7% of measurements within 5% error. OpenPose in frontal view performed poorly, with low correlation and high error margins. Processing time differed substantially between methods: MediaPipe averaged 331 seconds per video on modest hardware, whereas OpenPose required 2,612 seconds on a high-end system. Conclusion: These findings suggest that MediaPipe offers a more accurate, accessible, and efficient solution for finger-level ergonomic assessment in office environments. The results highlight the influence of camera view and algorithm choice in markerless motion capture, and support the future integration of AI-based tools into non-invasive ergonomic risk evaluation systems. This finding has implications for Realtime ergonomic risk monitoring. | ||
| کلیدواژهها | ||
| Artificial Intelligence؛ Biomechanical Movement Capture؛ Ergonomic Assessment؛ Finger Joints؛ Joint Range of Motion؛ Task Performance and Analysis | ||
|
آمار تعداد مشاهده مقاله: 3 |
||