By combining algorithms with sensors, engineers have created digital healthcare platforms to help the elderly age comfortably at home.
Nearly 20 years ago some of the world’s biggest technology companies tried to create digital healthcare platforms to help the elderly age comfortably at home. They had access to the brightest minds and deepest resources. What they didn’t have was a team of University of Missouri engineers and clinicians who beat them all to the finish line.
“This is why it’s so important to work with the right people to do the right thing,” said Marjorie Skubic, an electrical and computer engineering professor at the university, and one of the project leaders who worked on developing an automated, sensor-based network that helps the elderly live independently.
Over the years, companies like Intel, IBM, and Philips managed to develop stand-alone digital healthcare products, but they never succeeded in their ultimate goal of launching a fully integrated hardware and software solution. Skubic believes those companies were missing a key ingredient that helped her engineers succeed.
Combining Algorithms with Sensors
“When we first started out, a lot of people were developing this technology in a vacuum, without understanding the clinical problems,” said Skubic, who directs the university’s Center for Eldercare and Rehabilitation Technology, which grew out of the project. “From the beginning, this was an initiative that started from the clinical side. That’s what makes us unique.”
The university’s engineers and clinicians spent more than 10 years developing a platform that combines algorithms with sensors and connected devices to monitor and track changes in the activity and behavior of the elderly. The sensors range from infrared motion detectors to devices that measure heart rate, respiration, and restlessness during sleep. A 3D depth camera provides precise details on walking time, stride length, and the time it takes to move from sitting to standing, which algorithms use to predict the likelihood of a fall.
The information goes to healthcare workers, who look for warning signs of problems, such as lack of balance or loss of strength. They then intervene with treatments to improve health and keep people out of assisted living facilities and nursing homes.
The platform is now implemented in 25 apartments at TigerPlace, a 54-unit retirement community in Columbia, Mo., run by the university’s Sinclair School of Nursing and Americare, a healthcare services provider.
Seniors living with the platform stayed in their apartments 1.7 years longer than those without it, according to a recent study. In some cases, TigerPlace seniors without the platform moved to assisted living facilities or died from health issues that might have been prevented through early intervention.
“Our goal is not to increase the lifespan of individuals,” Skubic said. “We want to increase their healthspan.”
Tracking Gait-Related Data
One of the first projects Skubic’s team of nurses and engineers focused on was gait. Problems with walking are usually an early indicator of physical ailments such as cardiac problems, urinary tract infection, and acute pain, as well as such cognitive issues as depression and dementia.
Through trial and error, the team designed a sensor network that monitors how residents walked. A typical system for a one-bedroom apartment now includes a 3-D camera and image analysis software to record walking speed, stride time, and stride length. It detects falls and generates automated fall alerts with links to depth image videos that show what happened leading up to the fall.
Eight other sensors monitor and record motion where measurable activity is most likely to occur, like the front foyer, the bathroom, or the living room. A pneumatic bed sensor under the mattress captures pulse, respiration and restlessness at night, while a temperature sensor shows when someone is using the oven or stove.
The team used the data in two ways. First, healthcare workers created electronic health records that track falls, hospital visits, treatments, vital signs and other health data. Second, the team developed algorithms that classified the data in order to use it to predict future problems.
Creating the algorithms is one of the most difficult and time-consuming aspects of creating a platform of networked sensors, Skubic said. It is also one of the areas where most teams fail, mainly because they don’t work closely with experienced clinicians. “They know what the population needs,” Skubic said. “They tease the things out that are missing with engineers.”
To develop the gait algorithms, for example, physical therapists and nurses worked for months training and filming actors and others who mimicked the different ways elderly people walk, fall, sit, and stand. The team’s hard work enables it to predict the likelihood of falls.
“Interdisciplinary work gives value to what we do as engineers,” Skubic said. “We want to have impact of people’s lives, as opposed to just writing research a paper for a theoretical journal that no one would read.”
A Data-Centric Approach to Aging
One of Skubic’s former engineering students has launched a startup, Foresite Healthcare, to commercialize the technology. It has installed similar systems in 20 additional facilities, including three hospitals. Subscriptions to the service cost between $100 and $200 a month, depending on the application.
Skubic is not the only one taking a data-centric approach to aging. Ruzena Bajcsy of University of California, Berkeley, has been one of the nation’s top roboticists for decades. Now 83 and aware of her own frailty, she is developing models that will enable physicians and physical therapists to prescribe better assistive devices and customize rehabilitation exercises for patients.
That begins with taking kinematic and dynamic measurements, such as how far a person can reach, how they wash their face, or how they bend and tie their shoes.
This is more difficult than measuring the speed and length of someone's gait. “While you can measure a body in space with six degrees of freedom, a human body has many more,” Bajcsy explained. “If there is a problem with the knee, the body can compensate with other joints and muscles. This makes it difficult to detect which joints are weaker or stronger.”
Rather than go with expensive sensors, Bajcsy wants to use low-cost methods to capture human motion kinematics, such as cameras and motion sensors. She then wants to use the data they collect to characterize strength, flexibility, endurance, and balance.
“We want to create a system that produces an individualized model that changes as you do,” she said. This would enable professionals to evaluate the effectiveness of an exercise program or track the progression of a disease.
Ultimately, she hopes to build assistive devices customized for an individual’s specific capabilities. “We don’t want to give you too much torque and break your wrist,” she said.
Because Bajcsy and Skubic are basing their approach on data drawn from fundamental studies of human motion, that seems less and less likely.
Jeff O’Heir is a technology writer based in Huntington, New York.