Researchers crawl toward solution for cerebral palsy sufferers
BY THE NUMBERS
764,000 children and adults in the U.S. exhibit one or more symptoms of cerebral palsy
10,000 babies born in the U.S. each year, who develop cerebral palsy
Source: United Cerebral Palsy
1 in 303 estimated average of children in the U.S. with cerebral palsy
Source: Centers for Disease Control and Prevention
AT A GLANCE
Cerebral palsy is a collection of disorders caused by brain damage that affects the brain’s ability to send messages down to the muscles
Infants with a disorder that severely affects motor and cognitive development may learn to crawl on time thanks to a multidisciplinary group of OU researchers.
Andrew Fagg, Lei Ding, David Miller and Thubi Kolobe are combining robotics, brain imaging and machine learning to create a device that helps infants learn the specific body and limb movements required for crawling.
The team received a $1.135 million grant from the National Science Foundation in October to improve upon a device that Kolobe designed at Virginia Commonwealth University. Three years ago, Kolobe came to OU and presented the prototype to Fagg and Miller.
After meeting, Miller and Kolobe applied for and received a grant to develop a kinematic suit that observes infants’ limb movements in real time, said Miller, co-investigator for the project.
The suit fits the infant like a “onesie” and is functionally similar to body suits used in animation movie filming to simulate natural movement, Miller said. Twelve inertial measurement sensors spread throughout the suit communicate where the infant’s limbs are in space at any given time.
The new grant will be used to introduce brain imaging, Ding’s specialty, into the project, improve the kinematic suit and redesign Kolobe’s crawling robot, Fagg said.
Kolobe’s design for the crawling robot was a type of motorized skateboard, equipped with pressure sensors to read shifts in the infant’s weight and respond with movement accordingly, said Fagg, the project’s lead principal investigator.
The skateboard design, however, had some limitations, Miller said. It broke on a regular basis, and the device carried the infant’s whole weight, so the child never learned to support its own weight. The robot could not scoot sideways, a common movement for infants learning to crawl.
“We want something that can mimic all the movements that infants want to do and do the same type of training to get them to learn to carry more and more weight,” Miller said. “We’re pretty much starting from the ground up.”
The new device uses “omni wheels,” wheels with rollers perpendicular to the axle, that allow the device to skid across the floor. These wheels are similar to those used in factories to move parts across the production floor, Miller said.
A force-torque sensor that rests on the wheels measures forces that are up, down, left, right, forward or backward, including pitch and yaw, he said. The robot combines information from the sensor with information from the kinematic suit to generate motion in relation to the infant’s limb movements.
Infants learn to crawl by trying random movements, and these movements might cause them to roll or scoot across the floor, Fagg said. The resulting motion surprises the infant and drives them to keep trying.
“The infants start to build associations like ‘I see a toy a few feet away; here are the movements I need to get over there,’” Fagg said.
“Infants with cerebral palsy don’t have the strength or muscle coordination to produce those movements, and thus they don’t have those ‘Aha!’ moments. They learn to stop trying.”
The absence of these “Aha!” moments substantially delays crawling, and because crawling is key in childhood development, cerebral palsy infants face more developmental problems down the road, Fagg said.
Consequently, infants with the disorder have a hard time coordinating muscles and producing movement with enough force to induce motion.
The disease is non-progressive and does not worsen over time, but it has a dramatic effect on motor skills and cognition, Fagg said.
While infants with cerebral palsy do attempt motions that would otherwise cause the child to roll or scoot across the floor, they lack the muscle strength to initiate the motion. To compensate, the robot rewards the infant’s attempts by moving in a natural manner, Miller said.
The robot subsequently raises the bar and makes it slightly harder for the infant to initiate the robot’s motion, Miller said. For example, if movement of the right arm initially caused the robot to move to the right, the infant might have to move that arm while shifting to the right to cause the robot to repeat the same motion.
“At first if you wave your arms and legs, you’ll move; then we’ll fine-tune it,” Miller said. “By intermittingly rewarding simple motions, the infants eventually should be able to crawl.”
Several considerations go into making a robotics device for infants, Miller said. Everything must be soft and smooth, and the wheels must be guarded so the infants cannot run over themselves.
One of the biggest considerations, however, is getting the children in and out of the device quickly, he said.
“If it takes you five minutes to strap into the device, the infant’s going to be frustrated,” Miller said.
Electroencephalography (EEG) hairnets — devices that measure brain activity — will be placed on the infants while they learn to crawl on the robot, and Ding’s expertise in brain imaging will give the team an idea of what’s happening in the brain as the infants are developing, Fagg said.
The research team will work with healthy, developing infants in the first year of the grant and begin to work with cerebral palsy infants in early 2014, Miller said.
Toward the end of the grant, the team will begin to experiment with a robot that will help infants with cerebral palsy transition from crawling to walking, Miller said. They are dealing with very basic concepts now, but the team will apply for another grant if the designs look promising.