Technologyreview.com, April 14, 2004

Robo Rehab

Stroke victims recover more fully with the guidance of ever-smarter robots.

By Eric Bender

Use it-or lose it.

That familiar advice rings particularly true for survivors of a stroke. Each year 700,000 people in the United States have a stroke. About three-quarters survive, but more than half suffer from impaired movement. Their route to recovery is long and tough, as they painfully relearn how to use an arm or a leg by going through the motions over and over again with a physical or occupational therapist.

Unfortunately, all that therapist time gets very expensive, and health insurers have dramatically cut the amount of therapy they will reimburse. "When I started, it was three to five months," says Susan Fasoli, an MIT researcher and occupational therapist. "Now we're lucky if we get patients for three weeks."

Given less therapy, many stroke victims never recover as well as they might. "The more therapy you do and the more intense the therapy is, the better your ability to recover function after a stroke," says Richard Mahoney, president of Phybotics, a robotics startup in Westmont, NJ. But if patients can function-even one-handed-their insurance firms may tell them that their rehabilitation is done. And if training is overly focused on a few very specific tasks, it may even impede a more general recovery.

Enter rehabilitation robots, which can ease the therapist's load by delivering certain treatments very efficiently, in some cases, achieving dramatically better results than conventional therapy alone.

Researched for more than a decade, rehabilitation robots are being tested on patients throughout the United States, Europe and Asia. And they're just starting to appear in clinics for more general use. Interactive Motion Technologies in Cambridge, MA, has sold about two dozen systems for arm therapy, priced at $5,000 to $70,000, says company director Robert Parlow. Another market leader, Hocoma of Staefa, Switzerland, offers a robotic gait system, which combines with a treadmill to aid patients re-learning to walk.

Proving Patient Progress

Interactive Motion's robots are based on pioneering work by MIT researchers Neville Hogan and Hermano Igo Krebs. Optimized over the years, the basic design is a robotic arm that works in two dimensions and aids recovery of shoulder and elbow movement by carefully guiding the patient's partly paralyzed arm through appropriate motions, over and over. Patients watch a video screen and play "the world's most boring video game" with their disabled arm, says Hogan, a professor of mechanical engineering and of brain and cognitive sciences, as well as director of the MIT Arm Lab. The robot can exercise them more precisely than a human therapist, and it doesn't tire. "Within an hour, you can do about 1,000 repetitions of motion-much more than anyone would ever get within the context of a conventional therapist," says Fasoli.

According to Hogan, pilot studies have repeatedly demonstrated the value of the robotic approach, advocates say. Patients using the robot have shown twice the functional improvement, on standard clinical scales, as patients given conventional therapy, over the same treatment periods. And they continue to make progress in treatment programs months or years after the stroke.

"I think it's going to be a great adjunctive therapy," says Richard Hughes, a physical therapist at Spaulding Rehabilitation Hospital in Boston, which is doing research along with the MIT group. "Patients generally like the robot. Many think of it as similar to a video game." Using the robot in a highly structured way makes it easier for them to perform the motions-almost like patrons of a health club on a workout machine, he adds.

The MIT group is collaborating with rehabilitation clinics in pilot tests of new devices that add vertical motion and wrist movement capabilities to the robot arm. Results from early tests are encouraging, Hogan says. The lab is also working on systems for lower-limb recovery, and he envisions a family of machines.

In addition to precisely measuring limb movements and reacting accordingly to build the patient's coordination and strength, robots can modify treatments on the fly. Because of this flexibility, "we should be able to accelerate learning by the reward schedule," Hogan says. His lab is writing new algorithms that adapt the robot response to give better outcomes with fewer repetitions. "As you get better and better, the robot does less and less," he says. "You keep raising the bar. You don't want to discourage patients and you don't want them to go to sleep." Early tests show that this approach can triple the functional improvements of basic robotic therapy during a given treatment period, he says.

The U.S. Department of Veterans Affairs is a major funder of research into rehabilitation robots. Among other efforts at the VA's Palo Alto Rehabilitation Research and Development Center, H.F. Machiel Van der Loos and other investigators have pursued a dual-arm design that can help stroke survivors coordinate the movement of an injured arm with their other arm for combined tasks such as clapping. (Phybotics is working to commercialize this system.) Many other groups, both in the United States and overseas, are actively studying robotic rehab devices for upper or lower limbs, or both.

Some researchers are leveraging potentially complementary advances. Janis Daly, adjunct associate professor of neurology at Case Western Reserve's School of Medicine, is studying the effect on patients who combine conventional therapy with either robot therapy or with functional neuromuscular stimulation (FNS)-that is, the use of an electrical signal to directly activate a muscle. In preliminary results, patients with robotic therapy show good results for shoulder and arm coordination, while patients with FNS treatment demonstrate progress in wrist and finger control. "Our patients are so excited; they try to stay in treatment longer," Daly reports.

Arming the Brain

Throughout all of this work, many questions remain about the mechanisms by which the brain can rewire itself to recover motor control. While therapists have carefully gathered evidence to evaluate specific procedures, "they've never really done the work and explored the neural activity at the same time," says Phybotics's Mahoney. "The brain puts together all sorts of messages as to how to get your arm to move and pick up a cup," says MIT's Fasoli. "It's a very complex process, and we don't know a lot about it."

Among those attacking this problem, the MIT group has an ambitious plan to monitor patterns of brain activity via magnetic resonance imaging (MRI) while a patient works with the robot arm. That will be no small trick since, among other obstacles, an MRI device "essentially fries all electronics that are within range," Hogan notes. His lab is working on a nonmagnetic version of the arm that gathers data via optical sensors and is operated by hydraulics.

Proponents suggest other therapeutic uses of robotics beyond stroke recovery, including treatment of neurological conditions such as Parkinson's disease and cerebral palsy, sports medicine, and more general orthopedic rehabbing. And the advances in understanding how to build robots that interact safely with humans should pay off in personal robots and other devices, predicts Van der Loos of the Palo Alto research center.

As understanding grows and more robots enter rehabilitation clinics, experts are interested in developing versions for outside the clinic. While therapies will change over time, stroke victims will continue to get treatment for as short a period as possible until it is safe for them to go home, Fasoli says. With appropriate remote supervision from a well-staffed center, robotic devices might allow patients to continue intense therapy at home. In general, Veterans Affairs is "very proactive for this shift of responsibility from the hospital to the home environment," comments Van der Loos. "That's the way it's going to be; healthcare isn't getting any cheaper."

Another example of this trend into the home, Van der Loos says, is Java Therapy, a Web physical rehabilitation system created by David Reinkensmeyer and co-workers at the University of California, Irvine. Java Therapy offers exercises for stroke victims and others with motor impairments; its benefits will improve when its visitors are outfitted with robotic hardware that allows a wider range of activities.

But investigators emphasize that the robots will always need that human supervision. "Nobody doing this work is trying to replace the therapist; you can't," says Mahoney. "As robot devices get introduced to the clinic, therapists will carry out the same role but use the robotics as a tool."

"We're getting good feedback in terms of acceptance by patients and therapists, including therapists who had great skepticism," says Interactive Motion's Parlow. "It is literally and figuratively taking a load off them." Clinicians, he says, accept the robots more easily after deploying hardware for constraint-induced therapy-a popular approach in which an uninjured limb is restrained to help its impaired twin regain full function.

Tricky issues of reimbursement still need to be worked out with Medicare and other payers, Parlow says. But while his company's first systems all went to researchers, the devices now mostly end up at rehabilitation hospitals that are both carrying out research and considering clinical use. Very few other approaches have demonstrated clear, quantifiable benefits for stroke victims, he says. "The point of critical mass seems to be coming."