Fall 2012 Magazine

The Data Age

Prognosis: Better, Smarter, Faster

McCormick researchers are changing lives at the Rehabilitation Institute of Chicago


Download a PDF version of this story

Rehabilitation Institute of ChicagoIn a lab on the 14th floor of a downtown Chicago research center, Patrice Loudin sits in front of a computer. As researchers look on, a large red dot appears on the screen; with the aid of a robotic sling, Loudin slowly raises her right hand toward the target. While an eye-tracking device on her head shows that she’s aiming for the center of the circle, she manages only to graze its edge before slowly drawing her fingers back toward her lap.

Her aim may not be perfect, but the fact that Loudin is moving her arm at all is momentous: she is a quadriplegic. Using electrodes and a complex series of algorithms, this device-in-progress has provided the missing link between Loudin’s intentions and her movements—allowing her to control her arm for the first time in two years. “It feels natural,” Loudin says. “It feels like I’m using my arm naturally for the first time in a long time.”

Moments like these are the stuff of dreams for patients, doctors, and researchers at the Rehabilitation Institute of Chicago—and for the McCormick researchers at their sides. Named the top rehabilitation hospital in America by U.S. News & World Report for 21 years running, RIC is a leading provider of physical medicine and rehabilitation services, both at its 182-bed inpatient hospital in Chicago and at more than 50 satellite locations in the metro area. It’s also a hotbed for research, with its Searle Rehabilitation Research Center holding the title of the largest physical rehabilitation research center in the world.

A unique partnership between Northwestern and RIC has allowed several faculty members to hold joint appointments at McCormick and the Feinberg School of Medicine’s Department of Physical Medicine and Rehabilitation, housed at RIC. The collaborations bring together medical experts and therapists with researchers from biomedical engineering, mechanical engineering, and electrical engineering who work toward a shared goal: improving the lives of patients.

“It’s really nice to have these collaborations,” says Eric Perreault, associate professor of biomedical engineering and of physical medicine and rehabilitation. “I have engineering colleagues in Evanston whom I work with quite closely, and I’m able to work with patients and clinicians in a hospital setting here. It’s a nice synergistic relationship.”

New Hope for Paralyzed Patients

Eric PerreaultOne day the device Patrice Loudin is testing could help spinal cord injury patients regain control of their paralyzed limbs. In many ways the 19-year-old is the perfect candidate for the research: since suffering a high-spinal-cord injury two years ago, she has lacked motor control over her body from the neck down. She does have limited control of her neck and shoulder muscles, however, providing an opportunity for Perreault and his fellow researchers to help determine how Loudin intends for her arm to move.

“After a spinal cord injury, the electrical signals that control movement are blocked at the site of the injury and cannot get to the relevant muscles. But many of those signals can still be recorded,” Perreault explains. “By recovering signals related to the intended movement, such as in synergistic muscles above the injury, we can infer how the subject would like to move and then use our robot or another means to realize that intention.”

In essence, the goal is a device that can read a person’s mind. For the past two years Perreault and PhD candidate Elaine Corbett have pursued this goal, working with researchers from RIC’s Sensory Motor Performance Program, the Feinberg School of Medicine, and colleagues from Case Western Reserve University in Cleveland on a project funded by the National Science Foundation Program in Cyber-Physical Systems. They began by collecting data from 21 unimpaired subjects, developing a “user interface” that can translate thousands of intentions. Last summer they began using their algorithm with impaired patients.

An occupational therapist fits Patrice Loudin, a quadriplegic, with an eye-tracking device.During Loudin’s first laboratory session, researchers placed Band-Aid-like electrodes on her neck to monitor muscle activity and a reflective device over her eyes to follow her gaze. When a spot appeared on the screen, Loudin’s occupational therapist instructed her to try to reach for it, as she would have before her injury. As she did, the algorithm translated the activity in her neck and eyes, deciphering where she wanted to move her arm. The computer then sent direction to a robotic arm, which was attached to Loudin’s own arm, and it guided her hand to the desired location. In a perfect case, Loudin would have hit the circle in the center; that she just touched the edge of the circle could be due to weak signal reception in the electrodes, Perreault explained.

The ultimate goal of Perreault’s research is to combine the movement decoder with a fully implanted system capable of stimulating the paralyzed muscles so the intended movement is restored without the need for a robot. This technology, known as functional electrical stimulation, has seen some success at Case Western Reserve, which pioneered much work in this area. For example, the Freehand system developed there has allowed many patients with less severe impairments to regain control of their hands. “This has allowed many people to feed themselves and write, when they couldn’t before,” Perreault says. “It’s a dramatic improvement in quality
of life.”

The beauty of Perreault’s algorithm is its flexibility: it can be applied to a variety of subjects with different levels of injury. “Every injury is unique,” Perreault says. “This leads to each person with spinal cord injury having very different abilities. This algorithm can take those differences into account.”

The technology is still in the early stages, though, and it’s hard not to notice a tinge of disappointment in Loudin’s voice when she asks how long it will be before the completed system, including the implanted stimulator, is available for more than this hour-long laboratory test. A viable implanted solution, she’s told, is years away.

Smarter Assistive Devices

Brenna Argall with the robotic wheelchairIt may be missing its seat, but even in its current incomplete state, the automatic wheelchair whirring across Brenna Argall’s 1,000-square-foot lab might be the smartest wheelchair you’ve ever seen. Using a suite of infrared and ultrasonic sensors and two Microsoft Kinects—the same sensors used in the Xbox—this wheelchair can sense its surroundings, avoiding collision with objects in its path.

On its first run the wheelchair zips around the lab, navigating tight turns and avoiding walls and chairs like an oversized Roomba vacuum cleaner. Soon, Argall hopes, the chair will be able to do more, such as recommending alternate routes to avoid unsafe conditions—regardless of whether its rider has even noticed a problem. Moreover, the wheelchair will work with its owner to learn new behaviors, fine-tuning itself for his or her needs and wants.

“My interest is in making robots that can both learn and make decisions,” says Argall, June and Donald Brewer Junior Professor of Electrical Engineering and Computer Science and assistant professor of physical medicine and rehabilitation. “We’re talking about devices that can share control with the patient.”

Argall's team with robotic wheelchairArgall is Northwestern’s first joint appointment to RIC from the Department of Electrical Engineering and Computer Science; new to rehabilitation, she comes to McCormick from a background in robotics and machine learning. Her field is largely new terrain in rehabilitation; while robots are sometimes used in today’s rehab work, they serve primarily as tools for physical therapists. (For instance, KineAssist, a device created by two McCormick mechanical engineering professors, senses when a patient loses his balance and holds him upright.)

Argall envisions a marriage of rehabilitation and artificial intelligence that could make therapy more natural and user-friendly. She points to clinical research that indicates there is a one-year window following an injury when patients are willing to explore new coping mechanisms and control paradigms.

To maximize benefit in that short time, devices must be as intuitive as possible; too often, Argall says, even sophisticated assistive devices like complex electric prostheses are used simply for aesthetic purposes because the user is uninterested or unable to learn the control.

Argall and her team with the robotic wheelchair“In our lab, instead of the human needing to learn how to use the machine, the machine learns from and is trained by the patient,” Argall says. “After the year window, patients have figured out what works for them and are not as willing to learn something new, especially when relearning involves a machine that may not be intuitive.”

Using artificial intelligence with human subjects can be challenging because of a lack of guarantees on control. When using robots in the traditional low-level control scenarios of mechanical engineering, researchers can say with certainty how the device will behave under certain circumstances. When developing robots that can sense and respond to their environment using machine learning, certainty decreases, resulting in potential problems with human subjects and a slowing of the invention process. “This technology has huge potential because you can expand and generalize it in so many ways,” Argall says, “but we also have to work with these unknowns.”

Argall recently started working on the uncertainty problem with robotics expert Todd Murphey, associate professor of mechanical engineering. In the end, Argall and Murphey envision robot systems that can be taught by a human and still keep all of the formal control guarantees like stability. “We’re starting to build bridges between the machine-learning, data-driven background I offer and the formal control theory he offers,” she says, “seeing if we can find some middle ground.”

Deconstructing the Arm

While much of the progress at RIC can be seen with the naked eye, other work is more subtle. For some projects, years’ worth of data collection and careful study may be required before there is any contact with patients.

Wendy Murray (PhD ’07)That’s the case with research by Wendy Murray (PhD ’07), associate professor of biomedical engineering and physical medicine and rehabilitation, on conditions of the hand and arm. Since her days as a PhD candidate in McCormick’s Department of Biomedical Engineering, Murray has been using the principles of mechanical engineering to study the arm’s complicated mechanics: what each muscle does, how it attaches to the skeleton, how it interacts with the rest of the system. Years of unglamorous work gathering data from the Feinberg School of Medicine cadaver lab eventually yielded powerful results: a computational model of the human arm that incorporates data about every component into an adaptable model.

One of the conditions Murray studies is osteoarthritis, a degenerative joint disease that can make writing a note or turning the page of a book a painful task. Treatment sometimes involves surgery, but the procedure carries significant risks, and outcomes vary greatly from patient to patient. Murray, along with Jen Nichols, a PhD candidate in biomedical engineering, and Michael Bednar, a hand surgeon at Loyola University, is working to find out why. Are the variations caused by subtle differences in how surgery is performed? Or slight differences in the physiology of patients? If the latter, what muscle, bone, or ligament is the key?

The answers are elusive, in large part because the human arm and hand are so intricate. “If you look at illustrations of anatomy, you can see the upper limb is incredibly complex,” says Murray. “The way the joints move is complicated, the way the muscles attach is complicated. It’s this elegant, incredibly designed system, and from a purely quantitative sense we don’t know that much about how it works.”

Murray’s model—which was first created during her graduate work with advisers Scott Delp and Tom Buchanan (now at Stanford University and the University of Delaware, respectively) and further developed through Delp and Murray’s collaboration with Kate Saul during Murray’s tenure at the VA Palo Alto—provides a tool for understanding the workings of the arm in ways not possible with research on human patients. “We can simulate things in a model and then take it apart and say, ‘This is the muscle that is causing this,’” Murray says. “We have taken all this sophisticated information and put it in one place.”

Research that once relied on slowly collected data from numerous patients is now as easy as changing a few numbers in a simulation. With some minor tweaks, for instance, researchers can simulate a surgery on a patient weakened by osteoarthritis, eventually developing new surgical techniques targeted toward people with certain impairments. “You can do thousands of simulations, and patterns start to emerge,” Murray says. “Then we are able to take these results to clinicians like surgeons and physical therapists, and they start to say, ‘You know, we’ve seen something like that in the clinic.’”

Collaborating with McCormick colleagues has allowed Murray to branch into a new field of inquiry: how muscles interact with the brain. Working with Eric Perreault and other researchers, Murray studies the results of tendon-transfer surgeries. In these procedures, muscles are disconnected and reattached to give patients the ability to move a paralyzed limb; for example, a muscle from the elbow may be used to power the thumb. The success of tendon transfers relies not only on surgical procedures but also on the follow-up; patients must learn how to control their muscles in a new way. The results are mixed. While patients generally learn to use the limbs, they frequently lose range of movement. Interestingly, evidence indicates that this loss is due to failing communication from the nervous system instead of trauma to the muscle or bone.

“How does the brain suddenly adapt to the fact that when it wants to bend the elbow, it is moving the thumb? And how can we help people who have had this surgery use these limbs optimally?” Murray says. “Combining Eric [Perreault]’s and my complementary approaches to study the biomechanics and control of the upper limb, we can begin to bring these pieces together to help these patients access their own strength.”

Currently Murray is working with mechanical engineering’s Todd Murphey on a new project funded by the National Institutes of Health, investigating whether biomechanical modeling could help develop a controller for complex hand movements for prosthetics. “The thought is that if we embedded this model into a prosthetic hand, could the wearer make the prosthetic move just by thinking about how they used to move their hand?” she says. “The device could pick up signals, and the model could predict the movement that would have happened.”

The ambitious project brings together experts from Northwestern, the University of Colorado, and the Defense Advanced Research Projects Agency. As a first step, researchers collected electromyography data of different hand and arm movements and entered the data into their algorithm. The simulation was so complex it took three days to run.

If the hopes for such biomedical research can be realized, the impacts on the disabled and aging population would be profound. “There’s no simple solution, but this is a real issue our society has to deal with,” Murray says. “Helping people stay mobile is one of the best things you can do for their independence and health and happiness. And as engineers, we are really well suited to study movement. It’s physics, after all.”  

Undergraduates Design Products for RIC Patients

Design Thinking and Communication students (from left) Kashyap Saxena, Vinithra Rajagopalan, Hayley Blythe, and Cheyenne LynskySince 2003, McCormick undergraduates have also been engaged in a collaboration with RIC through a required two-quarter course, Design Thinking and Communication (formerly Engineering Design and Communication), in which freshmen design solutions for patients. The arrangement offers an excellent learning experience for students—many of whom are taking their first stab at design—while the students’ fresh perspective and outside-the-box thinking offer real solutions to RIC patients.

One of those patients is Joy Ray, who lost control of her right arm after suffering a stroke at the age of 33. Once-simple tasks like tying shoes and washing dishes were suddenly a source of frustration; zippers were especially problematic. “For the first 15 years after my stroke, I avoided zipper jackets entirely,” says Ray, now 59. “It’s a problem for so many people, and there’s not much on the market that can help,” she says.

Patient Joy RayLast spring a team of four undergraduates from the Design Thinking and Communication class met with Ray to discuss the problem. Her struggle to use a zipper with only one hand was clear, but the solution was anything but. At first the students planned a device that would clip to Ray’s pants, firmly holding one side of the zipper so Ray could focus on plugging the loose end. It didn’t work. “The sides of the jacket kept drifting apart,” says Vinithra Rajagopalan (biomedical engineering ’15). “We realized we needed a component that would hold them together.”

By the end of the quarter the design had changed considerably, into a metal U-shaped device with two clips to hold the jacket closed and keep the zipper taut. The device still ran into snags in a final unveiling with Ray: the fabric flap over the zipper kept getting in the way, and the design couldn’t easily be flipped to accommodate men’s jackets, which zip on the opposite side.

“It’s kind of nerve-wracking. The act of engaging a zipper has a lot more to it than we originally thought,” admits Cheyenne Lynsky (chemical engineering ’15). “The stress comes from not wanting to disappoint. We’ve tried really hard to make something that will make her life easier.”

Design Thinking and Communication students with their project, a treadmill for wheelchair athletesProblems are not always solved in one quarter. Many DTC projects span several quarters, with different teams contributing to a design before a client considers a problem solved. One long-term project was especially ambitious: a wheelchair training device for wheelchair athletes that could simulate an increase in elevation. McCormick students started last year by interviewing athletes and RIC staff to learn about the shortcomings of the institute’s current equipment. They heard that athletes were not able to get onto the existing trainer by themselves; two staff members had to lift them. Furthermore,
the trainer provided only one level of resistance and accommodated only one shape of athletic wheelchair, so many athletes couldn’t use it.

“We asked the class to tackle all these shortcomings and create a piece of exercise equipment that could be utilized much like a treadmill but by wheelchair users,” says RIC program specialist Eric Johnson.

Design Thinking and Communication student with wheelchair treadmillA first prototype created by students addressed all these concerns; it also featured a locking system to secure wheelchairs and a computerized control system to vary the strenuousness of workouts and to provide performance readouts. Later groups refined the design with a lower profile, allowing more independent accessibility, and a more secure lock to secure the wheelchair to the device.

In the end, three classes worked on the wheelchair trainer. Johnson says he hopes it will be added to the RIC facility in the near future. “Each of the teams worked to encompass every aspect of my vision in this device,” Johnson says. “It was a great experience for all of us.”  

By: Sarah Ostman