It’s not everyday that you see a six-month-old zipping across the floor on a motorized skateboard. Unless you are at the University of Oklahoma, that is.
There, a team of biomedical engineers and physical therapists have developed a motorized device to help infants at risk for cerebral palsy develop motor and cognitive skills. After a promising pilot study, the group began a larger trial of 45 infants this year. They use a robot equipped with power steering, a sophisticated machine-learning algorithm, and an endearing little cap studded with dozens of electrodes that track brain activity.
Cerebral palsy encompasses a number of early neurological disorders that affect movement and muscle coordination. The disorder can be caused by one of numerous insults early in life, including brain damage during birth, infection, and trauma. The condition is usually not diagnosed until the child reaches his or her first birthday, yet medical professionals agree that early intervention will often improve a child’s capabilities.
Early in her career at the University of Illinois in Chicago, Thubi Kolobe, a physical therapist and researcher, developed a movement test to predict which premature infants are most likely to develop CP or other developmental disorders. With the ability to identify those infants, she set out to intervene during a critical window of development (between 2 and 8 months of age), when a child begins kicking, moving and eventually crawling. The ability to move is not only important for the formation of motor skills, but also for brain development. As infants explore the world around them, they create new brain connections and develop cognitive skills such as spatial cognition, problem solving, and depth perception.
But if an infant tries to move and doesn’t get the desired effect—as occurs with many children with CP who do not crawl until age 2, if at all—the brain eventually prunes away those motor and spatial connections. And crawling tends to be a reward-based learning process: Without success, a child will eventually stop trying.
In 2003, Kolobe joined the faculty at OU and soon began collaborating with OU engineering professors Andrew Fagg, David Miller and Lei Ding to construct a device to help at-risk infants explore their environment and receive positive feedback that promotes brain development and encourages new motor skills. “As soon as you start to crawl, the world seems like a much bigger place,” says Fagg. “We hope, with the crawling, we’ll set them up to build other capabilities that will be really important later on in life.”
Now in its third iteration, the Self-Initiated Prone Progression Crawler or SIPPC (pronounced “sip-see”), is building on earlier elements: a soft pad that rolls along on three wheels and a sensor-studded outfit into which the infant is strapped. Fifty times per second, the 12 movement sensors would send measurements to a computer, depicting the child’s movement in 3-D on a nearby monitor. Additionally, cameras mounted on the SIPPC captured what each of the infant’s limbs was doing at a given time.
Perhaps more importantly, the movement sensor data was digested by a machine-learning algorithm that determined what the baby is trying to do, then directed the robot to give him or her an extra boost in the appropriate direction, rewarding the infant’s attempts. So if a baby girl tried to push off with her back feet, but without enough force to actually move, the robot would helpfully zip her a few inches ahead. If she moved an arm as if to push to the side, the robot could almost simultaneously turn her in that direction. Pilot tests with earlier versions of the device demonstrated that infants steadily engaged the robot to explore areas, and that those who wore the sensor-laden suit had more foot movement than infants who did not.
Funded by the National Science Foundation, the team has now begun the aforementioned study of 45 infants between the ages of 4 and 8 months. This time, SIPPC includes a cap dotted with dozens of small electrodes for detecting real-time EEG brain activity as the infants navigate. Right now, that brain data is being collected to see if there is a measurable change in brain activity over time as a result of using the device. Someday, however, it might even be used to create a brain-machine interface in which brainwaves alone predict an infant’s intention to move. “But that would be a long way off,” admits Fagg. “EEG is a very challenging domain, with very noisy data, and having an infant in motion dramatically increases the amount of noise.”
The study has another 6 to 9 months to go, and it will take time to get through the “overwhelming” amount of data produced by the experiment, says Fagg. Each session produces about 10 gigabytes of data, and the team just passed 1,000 sessions. “It’s wearing everybody down,” he says.
But there is light on the horizon. So far, the results seem to align with the pilot results, and parents are so encouraged by their children’s progress that they routinely ask to buy the device for home use. But the technology is not quite ready for that yet, says Fagg. “One needs to do a much larger-scale study with researchers at several universities,” he notes. “In the long run, we’d like to be there, but we need to do the science first.”