Precise quantification of eye-hand coordination (EHC) during pediatric dexterity tasks is limited by the lack of practical, high-resolution hand tracking methods suitable for children with brain injury or neurodegenerative disease. Traditional marker-based motion capture systems and instrumented gloves can interfere with natural grasp patterns and are often difficult to implement in clinical pediatric settings. We describe an adaptation of the Anipose markerless 3D pose estimation framework to enable synchronized three-dimensional hand kinematics and eye tracking during the Nine-Hole Peg Test (9HPT). The method integrates multi-camera video acquisition with task-specific neural network training optimized to detect fine finger movements across diverse pediatric hand sizes and grasp configurations. Camera placement and recording geometry were configured to reduce occlusion during peg manipulation and maintain multi-view visibility of hand landmarks. Model validation demonstrated low pixel error and stable three-dimensional reconstruction following confidence-based thresholding. The resulting workflow generates synchronized 2D and 3D visualizations, spatial coordinate outputs, reprojection-error metrics, and landmark confidence scores without requiring wearable sensors. This approach broadens the applicability of eye–hand coordination research within pediatric clinical populations and facilitates the development of more precise, quantitatively informed diagnostic assessments and targeted neurorehabilitation strategies for children with neurologic injury. Markerless multi-camera 3D reconstruction of pediatric hand kinematics during the 9HPT Integration of synchronized eye tracking and task-specific neural network training Output of validated 3D coordinates, confidence metrics, and visualization files suitable for clinical research
Rajkumar et al. (Fri,) studied this question.