Connecting Brain and Machine

Pauline Anderson

December 03, 2013

Bionic legs that use sensors and a control system to allow amputees to seamlessly traverse almost any terrain; robotic arms with a sophisticated brain-computer interface (BMI) allow paralyzed patients to closely match the speed and coordination of a typical human limb; even a computerized bladder that could eventually alert patients with spinal cord injury when to go to the bathroom.

These are just some examples of BMIs that harness electrical activity produced by neurons in the brain to control the movement of a variety of robotic devices. The hope is that in the not-too-distant future, patients with a variety of neurologic disorders may recover their mobility and leave their wheelchair and other clumsy assistive devices behind.

The field of robot-assisted healthcare is burgeoning, but although the technology is evolving rapidly, such issues as regulatory approvals, clinician training, and high costs stand in the way of biomedical robotics eventually becoming part of everyday medicine.

New research into robotic and neuroprosthetic technologies, along with several review and perspective articles examining the state of the art in this field, was highlighted November 6 in a special issue of Science Translational Medicine.

Robotic Legs

In 1 report, Michael Goldfarb, PhD, professor, mechanical engineering, and professor, physical medicine and rehabilitation, Vanderbilt University, Nashville, Tennessee, and colleagues describe components of the latest robotic leg technology. These components include ankle and knee motors, knee and ankle angle sensors, and heel and toe ground force reaction sensors.

The sensors replace aspects of the peripheral nervous system. Combined information from these sensors is fed into a microcontroller, which provides the equivalent function of the central nervous system (CNS).

To measure information from the CNS, and to act in unison with it, electrodes can be implanted in the peripheral nerves or motor cortex. Because the robotic limb is isolated from the metabolic power supply (the circulatory system), the prosthesis has its own power supply, often an electric battery.

Since the robotic prosthesis can emulate all aspects of muscular function, it can reproduce many biomechanical features that aren't possible with conventional prostheses. For example, users have enhanced gait symmetry and stable, controlled movements and can better negotiate slopes and stairs.

They're also less likely to fall. "[R]ecent studies indicate that the annual incidence of falls in the lower-limb amputee population exceeds that of the elderly population, the rate of seeking medical attention as a result of such falling is comparable with that of the institution-living elderly, and the incidence of falling (and consequently requiring medical attention) is higher in younger amputees than in older amputees, presumably because younger amputees are less restrained in their choice of activities and terrain," Dr. Goldfarb and colleagues write.

Another benefit of this new robotic leg is that unlike energetically passive prostheses, it doesn't necessitate compensatory movements that increase the stress on intact joints, which can lead to musculoskeletal degeneration.

The authors point out that studies on the biomechanical benefits of robotic leg prostheses with physical sensor interfaces have appeared in the literature and the devices have started to emerge on the commercial market.

Future models promise to be even more functional, and the authors expect that the full promise of robotic prostheses will increasingly be realized. The result, they said, should be improvement in patient mobility and quality of life.

Moving Arms

Similar translational technology is being applied to other limbs. Researchers have developed a robotic arm that patients with spinal cord injury and other paralyzed patients can learn to maneuver via a sophisticated brain-computer interface.

In one example, reported last December in The Lancet, surgeons using stereotactic image guidance with structural and functional MRI implanted 2 microelectrodes into the left motor cortex of a woman with chronic tetraplegia due to spinocerebellar degeneration. This allowed researchers to pinpoint and record neuronal activity when the woman was asked to imagine using her hand and arm.

After some practice, the woman was able to grasp items and fluidly move the hand with the coordination, skill, and speed of an able-bodied person.

In their current paper reviewing personalized neuroprosthetics in this issue, Grégoire Courtine, PhD, associate professor, Swiss Federal Institute of Technology Lausanne, and international paraplegic foundation chair in spinal cord repair, Center for Neuroprosthetics and the Brain Mind Institute, and his colleagues cite this and another example of leveraging neurons in the cortex to interpret a patient's intended motor action.

"This BMI enabled voluntary control of a robotic arm to perform seven-dimensional reach and grasp movements with remarkable fluidity and even safely bring a cup of coffee or chocolate from table to mouth — overcoming the technical and logistical challenges to bring a robotic system into the personal space," they write.

In an interview with Medscape Medical News, Dr. Courtine predicted that the first version of this prosthetic hand will be available within the next 5 to 10 years.

BrainGate System

Perfecting a robotic arm is part of an investigational device called the BrainGate neural interface system. As reported previously by Medscape Medical News, this system uses a sensor to monitor brain signals and uses computer software and hardware that turns these signals into digital commands for external devices. The sensor, a small patch of silicon containing 100 hair-thin electrodes that can record the activity of small groups of brain cells, is implanted into the motor cortex.

According to a dedicated Web site, the goal of BrainGate is to create technology that will allow severely disabled individuals — including those with traumatic spinal cord injury and loss of limbs — to communicate and control common everyday functions literally through thought. The group behind it includes leading academic institutions, corporations, and various nonprofit and government organizations who work on the research, science, and development of applied commercial technology.

Controlling 1 artificial arm with the mind is good, but controlling 2 is better. After all, coordinated hand movements are useful for such tasks as opening cans and typing on a keyboard.

To that end, researchers from Duke University in Durham, North Carolina, and from Switzerland and Brazil are using a BMI that allows monkeys to simultaneously control 2 virtual arms simply by imagining moving them.

Large-scale brain activity from a rhesus monkey was decoded and used to control reaching movements of both arms of a virtual monkey avatar toward spherical objects in virtual reality. Duke Center for Neuroimaging

In a new study also appearing in the dedicated issue, the researchers recorded nearly 500 neurons in the brains of 2 monkeys. With the help of a specific algorithm, they used the recorded brain activity to create a robotic link that rerouted motor commands to virtual arms.

The monkeys viewed avatar arms on a computer monitor and placed the virtual hands over square targets on the monitor. After 2 weeks of training, they realized they could control the virtual arms without moving their real hands by staring at the computer monitor. Eventually, researchers noted that the monkeys' brains started firing in response to the virtual arms as though they were real arms.

With this technology, which aims to bypass the spine, patients who are paralyzed could theoretically simply think about moving a limb to actually move it.

Wearable Robot

But it doesn't stop with arms and legs. Duke University neuroengineers are leading the Walk Again Project, a multinational collaborative effort to develop and implement the first BMI capable of restoring full mobility to patients with severe paralysis.

"This lofty goal will be achieved by building a neuroprosthetic device that uses a BMI as its core, allowing the patients to capture and use their own voluntary brain activity to control the movements of a full-body prosthetic device," according to the project's Web site. "This 'wearable robot,' also known as an 'exoskeleton,' will be designed to sustain and carry the patient's body according to his or her mental will."

The plan also involves creating a worldwide network of leading scientific and technological experts. "These world-renowned scholars will contribute key intellectual assets as well as provide a base for continued fundraising capitalization of the project, setting clear goals to establish fundamental advances toward restoring full mobility for patients in need."

Some new BMIs go beyond external virtual movements to tackle bodily functions. Patients with spinal cord injury can't feel or control their bladder because of the disrupted connection between the brain and the bladder's sensory and motor nerves. Enter an electronic device that replaces damaged nerves.

The device, currently being tested in rats, records sensory nerve information that helps decipher when the bladder is full and when it contracts. Recent experiments also showed that the surgically implanted device could successfully block and trigger bladder emptying on cue.

According to the United Kingdom researchers carrying out this work, the findings of these experiments represent a step toward developing BMIs that could help restore sensation as well as movement in patients with spinal cord injuries.

Other researchers in the field of neuroprosthetics are working with microelectronic retinal implant devices. Considerable progress has been made using electronic epiretinal or subretinal implants to restore vision to patients who are blind.

Such progress in retinal implants is discussed in a new paper appearing in this issue of Science Translational Medicine.

These devices have recently received market approval for treating patients with hereditary retinal degenerative diseases wherein retinal photoreceptor cells are lost but the inner retina and optic nerve are intact, said the authors, led by Professor E. Zrenner, chair, Department of Pathophysiology of Vision and Neuro-Ophthalmology, University of Tuebingen, Germany.

However, they said, many issues still have to be resolved. These include determining the best location to implant multielectrode arrays and the best ways to ensure the safety and longevity of the devices. Other unanswered questions include how to improve spatial and temporal resolution and whether color vision might be possible.

Softer, Smaller Robots

Although many surgical robotic tools are being developed, most are still fairly rigid, according to Robert Wood, PhD, professor, engineering and applied sciences, Harvard University, and Conor Walsh, PhD, assistant professor, mechanical engineering, Harvard University, Cambridge, Massachusetts. In an editorial in the special issue, they make the case for developing "softer" biomedical robots.

"[T]he human body is, other than bone, mostly soft," they write. "This means that rigid biomedical robots must rely on precise sensor feedback and high-performance control systems to ensure the safety of the patient."

A less rigid device could assist patients with physical or neurologic disorders, in part by matching natural body movements, and yet still apply substantial forces and torques when required, they said.

According to Dr. Courtine and his research associates, a field of "soft" neurorobotics is emerging that marries neuroprosthetics with soft robotics. Soft neurorobotics, they say, is a new class of rehabilitation interfaces that seeks to provide more natural, more interactive, and safer robotic assistance through soft design.

To this end, they said, engineers are using soft materials, such as silicon rubber derivatives, to design wearable interfaces that are comfortable and lightweight. They're also developing soft hardware, including stretchable sensors, and soft control algorithms that integrate multilevel biological and neurologic feedback to personalize the degree of assistance.

As well as being softer, future generation smart robots should also be smaller, especially if they're going to be used in the clinical setting, according to Dr. Wood and Dr. Walsh.

"Given the promise of robot-assisted health care, it is not surprising that there is a tremendous push in academia and the medical device industry to develop robots that are smaller, softer, and safer for use in clinical settings," they conclude.

For Dr. Courtine, an area where BMIs might have the greatest impact is in the field of rehabilitation, for example after a stroke, where the key to recovery is maintaining activity. He foresees the future rehab center using various brain-engaging robotic "tools" to stimulate muscles and boost plasticity.

But such approaches are complicated because they have to be personalized, with stimulation adjusted very precisely to the deficits of each patient.

"For example, every stroke is different, with different deficits, and for each patient, you would want to create a brain-machine interface for the right locations," said Dr. Courtine. "If you want to rehabilitate someone with a robot, you don't want this robot to be the robot of everybody else."

"Nontrivial" Hurdles

Today, neuroprosthetic interfaces remain mostly confined to sophisticated laboratories, Dr. Courtine and his colleagues point out. Getting these personalized devices to individual patients "is contingent on several nontrivial clinical, technical, organization and regulatory hurdles—the resolution of which relies on the concerted efforts of world-class engineers, clinicians, therapists, funding agencies and regulatory bodies."

One of the biggest dilemmas facing neuroengineers is creating "stable" and lasting neuronal recordings, Dr. Courtine told Medscape Medical News. "This is one area where we critically need to make improvements in order to reach clinical fruition," he said.

As it stands now, "after a few months, you start losing neurons one by one, so after 1 or 2 years, you have no neuronal recordings," said Dr. Courtine. Consequently, the patient needs to be regularly "reimplanted" with electrodes.

Regulatory pathways through the Food and Drug Administration (FDA) and the European Medicines Agency for the kind of multisystem designs upon which neuroprosthetic treatments rely are "prohibitively expensive" and complex, according to Dr. Courtine.

Others agree that regulatory approval for robotic prostheses is among the hurdles that must be overcome before these new technologies can reach clinical practice. As Dr. Goldfarb explains, a multijointed coordinated device like the robotic leg prosthesis is considered by the FDA to be a Class 2 device; thus, for it to reach the market, evidence will need to be established regarding its safety and efficacy.

Lack of training is another barrier. Physicians who prescribe prostheses aren't schooled in the field of robotics, said Dr. Goldfarb. The devices must be rendered usable for a clinician without robotics training, clinicians must receive such training, or the manufacturer "must provide some viable combination of device usability and clinician training," he and his colleagues write.

Yet another issue standing in the way of getting the technology into everyday practice is weak financial and business incentives, according to Nitish V. Thakor, PhD, director, Singapore Institute for Neurotechnology (SINAPSE), and professor, biomedical engineering, electrical and computer engineering, and neurology, Johns Hopkins University, Baltimore, Maryland, another contributor to the special issue in a Perspective article.

"This is an esoteric technology with a small market and few patients benefiting, and with a very complex interface between the technology and patient," Dr. Thakor told Medscape Medical News.

There's certainly no lack of research ideas and innovative technologies, but only a few "holistic solutions," added Dr. Thakor. "Up front industry-university partnerships would help, but are lacking in general," he concludes, partly because of lack of funding and issues related to confidentiality and intellectual property.

Sci Translation Med. November 6, 2013, issue.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.