Background: Virtual reality (VR) is increasingly used in stroke rehabilitation to remediate upper limb motor impairments. However, before the effectiveness of a VR-based intervention can be determined, it is crucial to know whether motor performance (endpoint trajectories) and the quality of movements made in a 2D virtual environment are similar to those performed in the physical world.
Objective: To compare kinematics of functional reaching movements performed in a low-cost 2D VR environment versus a comparable physical environment in post-stroke and healthy subjects.
Methods: Post-stroke hemiparetic (61.413.0yr) and healthy subjects (47.419.3yr) performed unilateral and bimanual reaching tasks in each environment. In the virtual environment, arm movements were tracked using the Kinect 2 camera (Microsoft, USA) and an avatar of the arm and hand was displayed on the screen. In both environments, arm and trunk kinematics were recorded using the Optotrak motion analysis system (27 markers; 120Hz) and 3D kinematics were reconstructed. Temporal and spatial characteristics of the endpoint trajectory and arm and trunk movement patterns were compared between environments and groups using repeated-measures analyses of variance. 
Results: Arm and hand orientation relative to object contact differed between environments and groups. Differences may be related to disparities in object affordances based on altered visual depth cues in the 2D virtual environment.  
Conclusion: Visual depth cues affect reaching kinematics and should be taken into account when designing and implementing 2D virtual environments for upper limb rehabilitation post-stroke.

REACHING KINEMATICS AND AFFORDANCES IN A 2D VIRTUAL ENVIRONMENT IN POST-STROKE PATIENTS.