Automatic posture analysis of dentists in real-world clinical settings via a composite OpenPose keypoint detection-geometric approach: a pilot study

This small pilot/internal feasibility study aimed to develop and preliminarily evaluate an automated composite OpenPose keypoint detection–geometric method for analyzing dentists’ sitting postures in real-world clinical settings. A questionnaire survey of 126 dental professionals identified high-risk musculoskeletal regions. Based on expert consensus and ergonomic guidelines, posture parameters and their normal reference ranges were defined. Five dentists were video-recorded during routine scaling procedures, and 30 representative image frames were selected. A custom algorithm was developed to automatically compute head–neck flexion angle, trunk flexion angle, knee joint angles, and a composite score quantifying deviations from ergonomic norms using OpenPose-derived keypoints. These automated measurements were compared with manually annotated ground-truth measurements using mean absolute error (MAE) and mean absolute percentage error (MAPE). The automated method demonstrated low mean absolute and relative differences with the manual reference measurements. The MAEs were 0.74° for head–neck flexion angle, 0.26° for trunk flexion angle, and 0.67° for knee joint angle—all well below the clinically acceptable threshold of 5°. The corresponding MAPEs of these three measurements were 1.11%, 2.94%, and 0.57%, respectively. And the composite score achieved an MAE of 0.84 and a MAPE of 1.96%. This study developed three posture parameters and a composite score as quantitative indicators of dentists’ postures in real-world clinical settings. It also developed a geometric algorithm based on two-dimensional keypoint detection to automatically extract these scores, achieving relatively high accuracy compared with manual reference measurements. Following further refinement and thorough validation, the composite score and geometric algorithm have the potential to be deployed in clinical settings to monitor dentists’ postures in real time and provide immediate alerts when poor posture is detected, thereby helping to improve ergonomic practices.