Eight synchronized GoPros hang from the ceiling of a Southwest Research Institute lab, recording SwRI engineer Travis Eliason as he jumps at the center of the lab space.
Within five minutes, software powered by artificial intelligence converts the footage into a skeletal version of Eliason, displayed on an adjacent computer screen.
The technology, known as ENABLE, has primarily been used in sports and performance settings, like assessing hamstring injury risk in college football players, optimizing baseball pitchers’ biomechanics and predicting injuries in Air Force special forces trainees.
Local applications include projects at Lackland Air Force Base, and working with the San Antonio Spurs and UT Health San Antonio’s new Center for Brain Health, among others.
The technology’s scope is increasingly expanding into medical settings, as researchers explore applications such as early detection of Alzheimer’s disease through gait analysis, as well as treating arthritis and limb rehabilitation.
“Doctors already use very simplistic movement biomarkers,” said Eliason, a lead engineer and biomechanist at SwRI who helped develop ENABLE, referencing tests like sit-to-stand or six-minute walk assessments commonly used in older populations. “If you can get a more comprehensive view of how someone is performing an activity, you can gain more detailed information and potentially uncover insights into their health status.”
The technology is also getting better.
SwRI scientists are seeking to bridge two separate areas of biomechanics to create more comprehensive software, one that would allow them to not only analyze and optimize movement, but also see what’s happening in our tissues and ligaments at the same time.
Markerless motion capture
Modern motion capture technology began taking shape in the 1970s and ’80s for research in biomechanics, kinesiology and orthopedics — as well as early computer-generated imagery in Hollywood and game studios.
Early systems required actors or athletes to wear suits dotted with physical sensors to track joint angles, a method that was expensive, time-consuming and restrictive, Eliason explained.
Markerless motion capture emerged in the early 2000s, gaining momentum with the rise of AI in the 2010s. Around 2017, SwRI researchers saw an opportunity to improve existing AI-powered systems like OpenPose, aiming to create a more precise technology for research and biomechanics.
“It went from $100,000 worth of specialized cameras in a specialized environment to GoPros anywhere you want to be,” Eliason said. “But that technology at the time wasn’t good enough for research. We set out to say … ‘can we make it better?’”
The ENABLE technology was the answer to that question, solidifying into its current form in the last five years.
In the SwRI lab, the eight GoPros hanging above a square platform with a force plate embedded in the floor capture Eliason as he jumps. Each camera’s video passes through an AI neural network, identifying 85 key points on the body.
Within minutes, a biomechanical model of Eliason appears on a nearby screen, providing data on joint angles, movement speed and force distribution.
One of the biggest advantages of the technology is its speed and portability. Researchers can bring the tech to gyms, basketball courts or military facilities, letting participants move naturally.
“We can now take the lab to the people,” Eliason said. “Recruiting someone went from a big ordeal to not a problem at all. And you’re measuring them in spaces where they’re already active. No extra burden.”
From the Spurs to special forces
SwRI researchers partnered with the Air Force at Lackland Air Force Base to gauge injury risk among airmen entering an eight-week special warfare preparatory course, which had a high injury rate.
More than 150 airmen were assessed on the first day of the course, looking at functional movements like squats, lunges and jumps, and then tracked throughout the full eight-week program. Researchers compared their baseline movement patterns with injury outcomes, identifying differences between trainees who remained healthy and those who were hurt.
Using those results, the team was able to build a predictive model that could flag individuals at higher risk. Based on their day-one movement patterns, “we were able to identify people who are four times as likely to get injured,” Eliason said. The results were published in a 2023 paper in Frontiers in Bioengineering and Biotechnology.
For baseball, SwRI researchers have used ENABLE to evaluate pitchers and test subtle changes that could reduce shoulder and arm injury risk without sacrificing pitching speed. Professional baseball teams use the technology as well, but the specific uses are kept under wraps.
“The easiest way to minimize injury is just throw slower,” Eliason said, “but don’t tell an athlete to throw slower.”
The software can model thousands of small adjustments to body mechanics and estimate which ones might reduce injury risk while preserving performance.
ENABLE has also been used to assess hamstring injury risk during sprinting among Emory University football players in Georgia.
The San Antonio Spurs also use the technology for sports performance optimization, and it will be active in the team’s new elite human performance center under construction at The Rock at La Cantera.
Movement as a biomarker
The technology is drawing growing interest beyond athletics, with clinicians and medical researchers beginning to explore how markerless motion capture could be used in health care and rehabilitation.
SwRI has an ongoing project with the Center for the Intrepid at Brooke Army Medical Center, which specializes in rehabilitation for wounded service members. The team is studying a carbon fiber device known as the IDEO, a brace-like system designed to help patients regain mobility after severe ankle or leg injuries.
The device was originally developed for injured soldiers whose limbs could be saved but who still faced long-term mobility challenges, Eliason said.
Now, researchers are testing whether the device could also help with tibial stress fractures, a common overuse injury among soldiers. Standard treatment typically involves weeks of reduced activity, which can cause trainees to lose conditioning and struggle to return to peak performance.
Using motion capture and biomechanical modeling, the research team is estimating how forces travel through the bone — with and without the device — to determine whether wearing the device could reduce stress on the tibia enough to allow patients to remain active while healing.
Researchers at UT Health San Antonio’s Center for Brain Health are also exploring the potential for motion capture to help detect neurological disease earlier than traditional screening tools.
The new center contains a long narrow room that utilizes the ENABLE software to look at changes in gait and movement, which research has shown can signal conditions such as Alzheimer’s disease and dementia many years before more severe symptoms show up.
More broadly, Eliason said, movement is increasingly looked at as an important biomarker in healthy aging, longevity and predictor of disease. “It’s not going to be a diagnosis,” Eliason said. “But it could be another tool in the doctor’s toolkit.”
The same idea could potentially be applied to detecting and managing arthritis, Eliason said. But such an application might require greater insight into what’s happening at the tissue level rather than movement and force patterns alone, what the researchers refer to as a “digital twin,” and where the technology is now heading.
Full-body modeling
The next iteration of the ENABLE technology is to combine the different layers of analysis into a model that captures the internal stresses and forces happening underneath the skin with the existing biomechanical data.
Dr. Lance Frazer, another biomechanist at SwRI who works on the technology, explained that at the most basic level, the tech generates a skeletal model from video. A second layer estimates joint forces and loads. The final layer goes deeper, using tissue-level modeling to estimate strain in cartilage, muscle and ligaments.
In the Emory College football study, for example, researchers found certain sprinting mechanics that can predispose athletes to hamstring injuries. But simply identifying risky movement patterns doesn’t tell a trainer or a researcher what’s happening inside the muscle tissue.
“The markerless system can say, ‘You sprint a certain way and that predisposes you,’” Frazer said. “But it doesn’t tell you the ‘why.’”
Instead of stopping at joint angles or speed, Frazer said, the modeling could predict how strain develops in specific muscles, ligaments or cartilage, and how close those tissues are to injury thresholds.
Frazer compared it to the kind of computer simulation used in bridge design, where engineers calculate stress on a structure before it fails. In biomechanics, the same idea can be applied to the human body.
That added level of granular detail would also be incredibly useful in medical applications. SwRI researchers are working toward the ability to have someone simply walk down a hallway and estimate the stress placed on their knee cartilage — information that could help predict arthritis risk years before symptoms become severe, Frazer said.
Individual components of the system already exist. The difficulty is linking them together into a coherent and useful model.
“The technology itself does not solve anything, it’s just a tool,” Eliason said. “But if you know how to use the technology, it’s only limited by your own imagination. It’s up to us and our partners to figure out how we can take this tool that we build and turn it into something useful.”
