In early 2013, I began working with Gregorij Kurillo, Jay Han and Ruzena Bajscy on the evaluation an interesting clinical use of the Microsoft Kinect Camera. This was a result of Jay’s experience as a physician in Physical Medicine & Rehab (PM&R), where he noticed as lack of affordable and accurate quantitative assessments for upper extremely impairments. Gregorij & Ruzena had previously done extensive research on developing software for the evaluation of the upper extremity based on motion capture technology. However, such technologies are quite expensive and difficult to maintain. The introduction of the Microsoft Kinect opened a new opportunity for use as a clinical tool for capturing upper extremity movements using low-cost and widely accessible technologies. Thus, Gregorij expanded his software to take inputs from the Kinect and we set about evaluating its accuracy and effectiveness as a clinical tool. The technology can not only help improve the effectiveness of physical therapy for patients with physical disabilities, but also bring clinical tools towards the home health care. This work was recently accepted in the Journal of Health & Technology. Below is an abstract of our work:
Background. In clinical evaluation of upper extremity impairments, there is a lack of assessment methods that are quantitative, reliable, and informative of the overall functional capability of an individual in an intuitive manner. We propose a new methodology for the assessment of upper extremity impairments based on the concept of 3-dimensional (3D) reachable workspace using Microsoft Kinect camera.
Methods. We quantify the reachable workspace by the relative surface area (RSA) representing the portion of the unit hemi-sphere that is covered by the hand movement. We examine accuracy of joint positions, joint angles, and reachable workspace computation between the Kinect and optical motion capture system. Furthermore, we report on validity and reliability of the RSA measurement.
Results. The results of our analysis on 10 healthy subjects showed that the accuracy of the joint positions was within 66.3 mm for our experimental protocol. We found that the dynamic angle measurements had relatively large deviations, ranging between 9˚ to 28˚. The acquired reachable workspace envelope showed high agreement between the two systems, with 95% of the RSA measurements deviating less than 5% of the total surface area of the frontal hemisphere. The Bland-Altman analysis in conjunction with high correlation coefficients (between 0.86 and 0.93) also showed a high repeatability between trials.
Conclusions. The findings indicate that the proposed Kinect-based 3D reachable workspace analysis provides sufficiently accurate and reliable results as compared to the motion capture system. The proposed method with its intuitive graphic representation of the individuals’ reachable workspace could be promising for various clinical applications for evaluating individuals with neurological or musculoskeletal conditions.