Motion Capture For Physical Therapy
To many physical therapists, motion capture (mocap) is an exotic concept. Yet, recent advances in imaging and computer vision have drastically changed the definition of mocap, and have finally made it practical for the clinic and athletic training. With these improvements, mocap could power up a new age in the field of physical therapy.
What does mocap mean historically? What does it mean today for physical therapy? What are the pros and cons of different technologies on the market?
What is the most practical solution for your clinic, and can mocap be practical to incorporate into your PT clinic? It depends. Let’s dig in!
Traditional marker-based multiple camera motion capture
Ease of use 1/5
Traditional marker-based mocap “rigs” are designed for high accuracy. They incorporate multiple cameras in a dedicated, precalibrated space and require the subject to wear a large number of markers, and potentially a mocap suit.
Traditional mocap systems were initially designed for Hollywood animation studios, with their high-fidelity requirements. Over time, these systems found their way to academic institutions and high end biomechanics research labs. They boast high definition and frame rates capable of capturing complex whole kinetic chain movements such as gait, or a professional baseball pitchers full speed movements. Research grade accuracy and the “gold standard” label enable them to find adoption in the scientific community and highly specialized, high end clinics.
Limitations of traditional motion capture systems are:
Near-prohibitive cost for most practices: most such systems cost $75,000 to $200,000. In some cases, costs can skyrocket past $1,000,000!
High operational requirements: marker setup, camera calibration, and software operation all take non-trivial amounts of time. Furthermore, complex post-processing, requiring dedicated, highly-skilled, and expensive staff, add more time and cost to an already operationally intensive process.
Difficult integration into standard clinic operations: Most capture runs with such systems take at least one (1) to one and a half (1.5) hours, and often the better part of a day. As a result, they pose significant disruption to standard clinic operations: a typical client session easily goes beyond 60 minutes.
Download Your Free Motion Capture for PT E-book Below!
Inertial measurement (IMU) & electromagnetic (EMU) motion capture
Accuracy 2/5 (edited from 3 to 2 based on feedback from IMU users)
Ease of use 3/5
IMUs and EMUs incorporate accelerometers, gyroscopes, and sometimes magnetometers into small wearable devices to extrapolate motion. IMUs and EMUs were first utilized in navigation systems for large commercial vehicles such as planes and submarines. Over the years made, they have made their way into consumer electronics for orientation tracking and were popularized by gaming systems such as the Wii. Within the context of physical health and wellness, IMUs and EMUs have become incorporated into some professional systems, but have found greater popularity in consumer-oriented products such as step trackers. With good calibration and operation, IMUs and EMUs can obtain excellent accuracy and capture rate. In some cases, they can sample at over 200Hz. Furthermore, most IMU and EMU systems are not constrained to a predetermined, precalibrated environment, and therefore can be used in many situations, both indoors and outdoors.
Limitations of marker based IMU / EMU systems are:
Long setup time: Each IMU / EMU needs to be precisely placed on specific parts of the body. If improperly placed, the resulting data may turn out to be meaningless, or worse: misleading.
Bad data due to marker slippage: markers can move around due to poor adhesion (e.g. sweat, friction) during movements and exercises, invalidating any collected data during that run.
Significant post-processing: resulting data needs to be manually analyzed and interpreted in a process which requires time and expertise.
Constraints on subject’s natural motion: the presence of markers on the subject’s body can affect the subject’s natural motion because of friction and weight of the attached sensors.
Some usage limitations: the presence of large metal objects or sources of electromagnetic waves can significantly alter the readings of some EMUs.
Multi-Camera Markerless Motion Capture
Ease of use 2/5
Multi-camera markerless motion capture systems are an extension of the original multi-camera marker based systems. These systems do away with the need for markers by leveraging advanced mathematical techniques and multiple 2D camera feeds to identify body positions in real time at a high frame rate. They were initially designed to enable more affordable animation capabilities for movie creators and video game designers, but have found low penetration in the physical wellness space. The accuracy of such systems is good but still a step below that of marker-based multi-camera systems. Calibration and setup processes are still required. Such systems are not typically portable, require dedicated space, and can cost over $60k. The convenience of no markers is somewhat offset by the reduced accuracy, significant cost, and space and calibration requirements.
Limitations of markerless multi-camera motion capture systems are:
High cost for most practices: most such systems cost over $60,000.
High operational requirements: while marker setup is avoided, a calibration and setup phase is still required. Additionally, non-trivial post-processing is required to obtain relevant biomechanical analysis.
Lastly, such systems are not portable and require significant dedicated space.
Download Your Free Motion Capture for PT E-book Below!
Single camera 3D markerless motion capture
Ease of use 5/5
Single camera 3D markerless motion capture systems leverage the latest advances in machine learning and computer vision techniques to provide good body tracking at a fraction of the cost and setup required of the above listed systems. Single camera markerless 3D systems were originally created for general human recognition, but today, several of these systems have been designed specifically for the practicing clinician, from the ground up.
Because of the ease of use of these systems, movement screens can be conducted in 1-2 minutes. Post-processing is not required, as biomechanical analysis software is typically prepackaged. These systems are lightweight and portable and can conduct full kinetic chain analysis at the push of a button. An example of a 3D markerless motion capture is EuMotus BodyWatch.
Limitations of single camera 3D markerless motion capture systems are:
Situational constraints: some computer vision models require the subject to be facing the camera to properly function. This eliminates movements in the prone position (e.g. push ups, planks etc). Some of these systems also rely on infrared technology, which can be impacted by natural sunlight and lose accuracy in certain poor lighting indoor or outdoor situations.
Reliance on implied skeleton models: machine learning and computer vision techniques construct a probabilistic skeleton model from a camera feed without the need of markers. The results in a lightweight and easy-to-use system. The trade-off for this convenience is a reduced accuracy compared to traditional mocap. White papers show 3 – 9 degrees difference vs gold standard traditional mocap systems.
Lower frame rate: because of the intensity of the required image processing, the frame rate of such 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. full speed baseball pitch, tennis serve – will be out of scope for these systems.
2D video playback + annotation
Ease of use 4/5
2D Video playback systems have been around for many years as a cheap way to do video analysis for a large range of use cases. Today, with the proliferation of cell phone cameras, this type of software can be installed on a phone or tablet as an app. It’s easy to set up and run and can be used outdoors. No laptop or computer is required given the minimal amount of data processing.
Within the context of biomechanical analysis, several applications have been created to enable manual annotation and calculation of biomechanical angles. Going back and estimating an angle from a still frame may work better than a just-in-time naked eye observation in controlled scenarios.
Manual post-processing: most 2D video playback systems require human intervention to manually delineate the joints and angles of interest.
Inaccurate and sometimes misleading calculations: most measurements obtained by drawing on 2D video still frames are incorrect unless the camera is directly perpendicular to and centered on the angle of interest at all times. In most use cases, that requirement is not followed nor practical and results in erroneous measurements. 2D video analysis is considered a qualitative, not a quantitative analysis system.
Not holistic: 2D video analysis is typically limited to specific frames or focuses on single joints. It is not effective in capturing and analyzing full body dynamics.
Different end users have different requirements. The choice of a mocap system therefore is a unique decision making process.
A traditional mocap system may satisfy the needs of an academic biomechanics lab and may be out of reach for a one-off PT clinic. It can however, be the right choice for your clinic, and a major differentiating factor. With a dedicated space, dedicated resources and expertise, a multi-camera system could be set up at a high clinician / patient clinic in a location where there is market demand.
For clinics focusing on a high velocity data capture (e.g. golf swing), IMUs or EMUs may be a good way to go. For a locale with a large supply of golf enthusiasts or higher levels of baseball players, this may be a great way to cater services to the local sports community.
For most clinics, markerless motion capture is an interesting way to add value. Designed for ease of use and the clinician in mind, these systems can screen patients in minutes and can be integrated into the clinic operational flow. They can be used for return to sport, as well as for systematic screening of teams to find individuals and/or groups displaying faulty movement patterns.
Finally, 2D smartphone mocap apps are simple ways to record motion, and are one of the few (if not only) way to record motion at close range outdoors. Biomechanical analysis functionality is limited at best, yet while these systems lack accuracy and vast majority of analytics provided by other systems, they can be used to occasionally capture gross faulty movement patterns (e.g. valgus, at great angle). These type of systems might be great for your middle or high school soccer or basketball coach.
Considering integrating mocap into your clinic? Send us a note at hello/at/eumotus/com!
Download Your Free Motion Capture for PT E-book Below!
 1979 revolution motion capture. Wikimedia. https://upload.wikimedia.org/wikipedia/commons/f/f4/1979_Revolution_motion_capture_1.jpg.
 Nintendo Wii. Wikimedia. https://upload.wikimedia.org/wikipedia/commons/3/3e/Wiimote-in-Hands.jpg.
 Microsot Kinect. Flickr. https://c2.staticflickr.com/8/7457/14175572202_64b7e17e59_b.jpg.
 Cell phone. Pixabay. https://pixabay.com/p-1976104/?no_redirect.