Last week our physiotherapist in residence and chief scientist, Merci Greenaway presented the world’s first continuing education course on 3D markerless motion capture for the clinic. In the presentation, she explored how technology can be harnessed to further the field of Physical Therapy globally. Here we have a summary of the course, which can be found in its entirety on the physicaltherapy.com web platform.
Course Description: This course will provide an overview of 3D Markerless Motion Analysis as an objective outcome measure in the clinical setting. The evidence base for these tools will be discussed, as well as their clinical viability. Now when we say clinical viability we mean how realistic is it to use a particular motion analysis system in the clinical setting? Factors such as time efficiency, cost, and ease of use will be considered. We will also look at whether the information yielded from various systems is accurate enough to be used with confidence to make clinical decisions.
An outcome measure is any characteristic or quality measured to assess a patient’s status (Fetters & Tilson, 2012). An outcome measure is the result of a test that is used to objectively determine the baseline function of a patient at the beginning of treatment. Once treatment has commenced, the same instrument can be used to determine progress and treatment efficacy.
With the move towards Evidence Based Practice in the health sciences, objective measures of outcome are important to provide credible and reliable justification for treatment. Both for your patients and for their insurers. Now that we have some form of direct access in all states of the US, this may be the next step for PT. Namely, using technology to generate patient demand, prove PT results in superior health outcomes relative to alternatives (e.g. surgery, or opioids etc.) , and prove that PT is money well spent.
At the moment there is a push toward value-based healthcare. As we are shifting form fee per service to this value-based reimbursement it is important to be equipped with tools that show your value. Increasing, or conversely decreasing revenue by tens of percentage points is a strong financial motivator to generate objective data that demonstrates the effectiveness of your clinic to the payor and other interested parties.
From a business perspective, people often go to a physical therapist when they are hurt, why don’t we have annual check-ups with physical therapists? Just like you’d go to the dentist. Outcome measures are a great way to provide a baseline for people during an annual neuromusculoskeletal check up that Physical Therapy should absolutely be an equally valued preventative healthcare solution and outcome measures are a great way to prove that.
In the 1970s motion capture (MOCAP) was proposed as a method for improving animation. MOCAP has been used since the 1980s for animation and filmmaking, virtual reality, and gaming (Yang, 2015). From initial adoption by Hollywood studios, motion capture has since been adopted by university biomechanics labs and is gaining more and more adoption in the clinical space.
Early on, a lot of equipment and highly-specialized labor was required to extract meaningful information out of traditional motion capture multi-camera, marker-based systems. While traditional motion capture units remain operationally and resource intensive, recent advances in hardware (e.g. 3D cameras), and increased accessibility to powerful algorithms (e.g. computer vision, machine learning), motion capture and analysis technology have finally become accessible for the PT clinic.
CLINICAL VIABILITY OF MOTION CAPTURE SYSTEMS
EXAMPLE COMPARISON: UNAIDED VISUAL OBSERVATION
Unaided visual observation is the standard for physical therapy examination. It is quick, and does not cost anything. Benefits include that it is free, and easy to use for single body joint measurements. While visually observing a measurement is easy to perform, researchers have found inferior accuracy when compared with goniometry or motion capture tools (Hayes, Walton, Szomor, & Murrell, 2001; Watkins, Riddle, & Lamb, 1991).
However, it is difficult to assess multiple joints at a time: which is particularly important when assessing functional movements which involve multiple joints and body areas. Visual observation has been shown to have the greatest degree of error among the assessment tools discussed. As discussed, we use these measurements to determine the progress of a patient and to make decisions about treatment so, it is certainly a very important point that visual observation is not particularly accurate. Because this ultimately can affect patient outcomes
As described so well by, Colyer, Evans, Cosker, & Salo, (2018), “Clinicians want a motion analysis system that collects accurate kinematic data in a timely, unobtrusive and externally valid manner.”. This problem statement comes back to the idea that clinicians want a motion analysis system that they can practically and confidently use in the clinic.
One solution to the problem of balancing accuracy with ease of use is single camera markerless motion analysis, like BodyWatch by EuMotus. The key enabler for a markerless MOCAP system in the clinical context is the speed of setup, the speed of operations, and minimization/ elimination of post-processing. We designed our clinical motion analysis with the constraint that time is critical in the clinical setting. Many movement screens can be conducted in as little as one to two minutes. Post-processing is not required, as our biomechanical analysis algorithms immediately and automatically process raw movement data. This design not only saves time but crucially enables motion analysis for use in the clinical setting. Full kinetic chain analysis can be conducted at the push of a button.
3D markerless motion capture and analysis systems are lightweight and portable. Physical therapists can be confident with the results provided - 3D markerless motion capture scores well against gold standard biomechanics lab traditional MOCAP systems. Furthermore, markerless MOCAP systems serve as a tool to verify the clinician’s intuition, as markerless MOCAP is more accurate, more repeatable, and better suited to systematically measure multiple joint measurements and AROM at the same time than unaided visual observation, or goniometry.
HOW IT WORKS
The first element for a system like BodyWatch is a standard color image camera (Khosrow-Pour, 2017). This is an RGB color camera, similar to your cell phone’s camera. What makes these 3D cameras different is that they also use depth-sensing technology and computer vision. The system sends out infrared light and times how long the refracted light takes to come back to the sensor, yielding a 3D image. Computer vision algorithms are trained to recognize the shape of the human, and the human’s body joints.
EuMotus BodyWatch computer vision and machine learning algorithms eliminate the need for manually drawing lines. Our system automatically recognizes the person and estimates with good reliability where each body joint is based on computer modeling.
While, in my opinion, markerless motion capture systems strike the best balance of clinical utility and accuracy, it is important to discuss systems limitations.
Situational constraints: our system requires the subject to be facing the camera to properly function. This eliminates movements in the prone position (e.g. push ups, planks etc).Systems relying on infrared technology can be impacted by natural sunlight, or low light conditions and lose accuracy in certain outdoor or indoor situations.
Reliance on implied skeleton models: machine learning and computer vision techniques construct a probabilistic skeleton model from a camera feed without the need for markers. The results in a lightweight and easy-to-use system. The tradeoff for this convenience is a reduced accuracy compared to traditional MOCAP, however, traditional MOCAP is not easily integrated into the clinical context. When compared with the status quo for movement assessment, which is unaided visual observation, markerless motion capture systems strike the best balance of clinical utility and accuracy.
Lower frame rate: because of hardware limitations, the frame rate of single camera 3D infrared systems hovers around 30 frames/second. While this is sufficient to analyze most movements, such as gait and jumps, certain specialized and fast movements – e.g. baseball swing, tennis serve – will be out of the scope of these systems. The frame capture rate has improved and we expect it to keep improving with time.
WHY MOVEMENT ANALYSIS?
So why should you implement a motion analysis system at your clinic? There are two key factors to consider: patient engagement and clinic differentiation. There is good evidence for improved health outcomes because of patient engagement using technology (Filippeschi et al., 2017).
From the business perspective, a motion capture tool can be used to provide new services to cater to current, or perhaps new groups of customers. Empowering the clinician with a tool that can be used to educate the patient, better engage them, and better retain them as a customer is a strong reason to consider this product for your clinic.
Physical therapy clinics looking to differentiate themselves to potential clients and referring physicians can use motion analysis to demonstrate their commitment to clinical excellence. By applying motion analysis technology, patient’s recovery can be objectively tracked, proving the value of your interventions, and allowing yourself to truly differentiate your clinic.
Bakhshi, S., Mahoor, M. H., & Davidson, B. S. (2011). Development of a body joint angle measurement system using IMU sensors. In 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society (pp. 6923–6926). https://doi.org/10.1109/IEMBS.2011.6091743
Cloete, T., & Scheffer, C. (2008). Benchmarking of a full-body inertial motion capture system for clinical gait analysis. In 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (pp. 4579–4582). https://doi.org/10.1109/IEMBS.2008.4650232
Colyer, S. L., Evans, M., Cosker, D. P., & Salo, A. I. T. (2018). A Review of the Evolution of Vision-Based Motion Analysis and the Integration of Advanced Computer Vision Methods Towards Developing a Markerless System. Sports Medicine - Open, 4(1), 24. https://doi.org/10.1186/s40798-018-0139-y
Dockstader, S. L., & Murat Tekalp, A. (2001). Multiple camera tracking of interacting and occluded human motion. Proceedings of the IEEE, 89(10), 1441–1455. https://doi.org/10.1109/5.959340
Fetters, L., & Tilson, J. (2012). Evidence based physical therapy. FA Davis
Filippeschi, A., Schmitz, N., Miezal, M., Bleser, G., Ruffaldi, E., & Stricker, D. (2017). Survey of Motion Tracking Methods Based on Inertial Sensors: A Focus on Upper Limb Human Motion. Sensors, 17(6), 1257. https://doi.org/10.3390/s17061257
Finkelstein J, Knight A, Marinopoulos S, et al. Enabling patient-centered care through health information technology. Evid Rep Technol Assess (Full Rep). 2012;(206):1-1531.
Geerse, D. J., Coolen, B. H., & Roerdink, M. (2015). Kinematic Validation of a Multi-Kinect v2 Instrumented 10-Meter Walkway for Quantitative Gait Assessments, 1–15. https://doi.org/10.1371/journal.pone.0139913
Gray, A. D., Willis, B. W., Skubic, M., Huo, Z., Razu, S., Sherman, S. L., & Guess, T. M. (2017). Portable and Inexpensive Tool to Using the Microsoft Kinect V2, 9(6), 537–544. https://doi.org/10.1177/1941738117726323
Khosrow-Pour, M. (2017). Encyclopedia of Information Science and Technology, Fourth Edn. Hershey, PA: IGI Global
Lee, S. H., Yoon, C., Chung, S. G., Kim, H. C., Kwak, Y., Park, H. W., & Kim, K. (2015). Measurement of shoulder range of motion in patients with adhesive capsulitis using a Kinect. PLoS ONE, 10(6). https://doi.org/10.1371/journal.pone.0129398
Miezal, M., Taetz, B., & Bleser, G. (2016). On inertial body tracking in the presence of model calibration errors. Sensors (Switzerland), 16(7), 1–34. https://doi.org/10.3390/s16071132
Ministry of Defence, C&C, Defence Medical Services (DMS) (2017) Procurement of Military Performance Analysis and Research Laboratory (MPARL). Voluntary ex ante transparency notice Supplies
Seaman A., & McPhee, J. (2012). Comparison of optical and inertial tracking of full golf swings. Procedia Engineering, 34, 461–466. https://doi.org/10.1016/j.proeng.2012.04.079
Sigal, L. (2012). Lecture 3 ( Marker-based ) Motion Capture, 3
Watkins, M., Riddle, D., & Lamb, R. (1991). Reliability of goniometric and visual estimates of knee range of motion obtained in a clinical setting. Physical Therapy, 71, 90–97
Yang, L. (2015). 3D Sensing and Tracking of Human Gait by
Zhou, H., & Hu, H. (2008). Human motion tracking for rehabilitation-A survey. Biomedical Signal Processing and Control, 3(1), 1–18. https://doi.org/10.1016/j.bspc.2007.09.001