Motion Capture of Bronchoscopist Kinematics Using a Spatial Computing Device: A Methodological Proof-of-Concept Study.

Abstract

Introduction: Bronchoscopy is physically demanding, and studies of ergonomics and objective skill assessment require accurate, segment-resolved capture of the bronchoscopist's hand and head movements. However, existing motion capture approaches typically rely on specialized laboratory infrastructure or wearable sensor arrays, limiting their routine use.

Methods: A board-certified pulmonologist performed 10 systematic bronchoscopies across the 18 named segments using an airway simulator and a single-use bronchoscope. Procedures alternated between conditions without and with engagement of the bronchoscope's insertion-tube rotation mechanism. A custom visionOS application running on Apple Vision Pro captured bilateral hand and head poses at 30 Hz, with synchronized voice-cue annotations for each segment.

Results: All 180 expected segment insertions were recorded (94.4% via voice cues; 5.6% recovered through tracking-derived hold detection and excluded from per-insertion analyses). Intermittent tracking-freezing artifacts, in which a static pose was retained under a valid tracking flag, were detected by frame-to-frame immobility and excluded from the affected analyses. The mean per-insertion peak left-wrist rotation across all segments was 107.9° without and 72.6° with the rotation mechanism. The mean left-thumb path length within the wrist's local reference frame was 3.82 ± 0.48 m per bronchoscopy, and the estimated number of bronchoscope lever operation cycles was 64.8 ± 7.9 per procedure.

Conclusion: In this proof-of-concept study, a single off-the-shelf head-mounted device enabled multichannel capture of bronchoscopist kinematics during simulated bronchoscopy. This approach provides a practical methodological foundation for future studies on bronchoscopy ergonomics, skill assessment, and AI-driven training data generation.