It's the stuff of science fiction: a bionic arm that slips onto the stub of an amputated limb and supplies all the movement and dexterity of the real thing. Movie characters since the days of Luke Skywalker have been attaching them, gazing hopefully, then wiggling their fingers in a modified victory dance to show that their artificial limbs were every bit as good as nature's original.
Film fancy? Maybe not.
At the Rehabilitation Institute of Chicago, physicians earlier this year laid the foundation for such technology when they strapped Jesse Sullivan's new arm and shoulder into place. Sullivan, a Tennessee power company lineman whose arms were amputated after he was electrocuted on the job, suddenly had an artificial limb that allowed him to rotate his wrist and upper arm, bend his elbow, grip with his hand, and, incredibly, feel. And like Luke Skywalker, Sullivan required just a few minutes of learning to adjust to the new limb, mainly because it used a neural wiring scheme similar to nature's own.
"What's interesting about Jesse is that he simply does what he always did prior to the accident," notes Bill Hanson, president of Liberating Technologies, Inc., the company that made Sullivan's digital arm. "If he wants to lift something off a table, or reach for something over his head, his thought process is exactly like anyone else's." Hanson adds that because the prosthesis is so schematically similar to a real arm, Sullivan learned how to use it in about 90 minutes.
"There's a lot happening under the hood here," adds Kevin Englehart, associate director of the Institute of Biomedical Engineering at the University of New Brunswick and a contributor to the design. "But for the user to learn, it requires just a quick calibration."
Re-wiring the Body
Indeed, a great deal is happening "under the hood" of Sullivan's prosthesis, but the ones who most effectively grasp its advantages are those who have used conventional prostheses. Even the best of today's powered versions can't execute the direct brain-to-hand technique employed in Sullivan's unit. Rather, they require
Sullivan's arm-and-shoulder system includes motors for the elbox, shoulder, humerus, wrist and hand.
users to relearn the simplest motions. People who have lost a hand, for example, might use their biceps or triceps muscle to initiate powered hand movement. Moreover, many such prostheses can't execute a quick series of motions, requiring instead that users step through a sequential parade of muscle contractions to, say, simultaneously reach and turn.
"When you re-wire the arm so that the biceps controls the hand, it takes a lot of re-learning for the user," Hanson says.
Indeed, that convoluted re-wiring process was, at least in part, the motivation behind Sullivan's new arm. Todd Kuiken, a physician who holds a Ph.D. in biomedical engineering, first conjured up the concept while reading a medical journal more than 20 years ago. His idea was to use so-called "nerve reinnervation" to gain additional control signals for a prosthesis. That way, the prosthesis could take advantage of existing nerves, even if the limb associated with those nerves had already been amputated.
While there's a simple beauty to that concept, Kuiken learned over the years that connecting nerves directly to prosthetic devices was unrealistic. Signals in the human nervous system, he found, were far too weak to control such devices. To solve the problem, Kuiken decided to connect the nerves instead to usable bands of muscle, which could serve as amplifiers for the weak signals coming from the nerves.
As best as anyone can tell, Kuiken's method is unique in the annals of medicine, and it is opening up possibilities unlike any in the recent history of prosthetics. Today, he is combining his reinnervation scheme with myoelectric sensor technology and so-called "digital limbs" to enable Sullivan to accomplish tasks that would have been impossible in the past.
"With this arm, Jesse can take off his baseball cap and put it back on," says Kuiken, who serves as the director of amputee programs and associate dean at the Rehabilitation Institute of Chicago. "He can reach up and take things from a cupboard. He can grasp and turn a doorknob."
Moreover, Sullivan can do all those tasks, especially complicated ones such as grasp-and-turn, with merely a vague notion of desire. His conscious thoughts travel from his brain, through his nerves, and into a digital arm that interprets them.
Incredibly, Sullivan is also regaining a sense of feeling. "If you touch his chest, he feels it in his 'hand,'" Kuiken says. "Ultimately, that gives us a portal to give him true sensory feedback. We're working on the idea of him being able to feel what he is squeezing."
Motors That Respond to Thoughts
Making all of that happen, however, has been a long, laborious task for Kuiken and other researchers around the world, as well as for contributing suppliers, such as Liberating Technologies. The latest version of Sullivan's arm, outfitted for him this past February, contains a motorized elbow, shoulder, wrist, humerus, and hand.
Boston Arm's 2x6-inch printed circuit board incorporates six drives for the arm's six motors
In all, the system uses six motors, including one brushless DC unit in the elbow built by Liberating Technologies, two in an artificial hand created by the Keshing Co. of China, one motor in the wrist rotator built by Otto Bock Healthcare (Germany), one brush-type motor in the humeral rotator constructed by Northwestern University, and one in the artificial shoulder joint designed by engineers at the University of Strathclyde (Scotland).
All of the motors respond directly to Sullivan's thoughts. The device's operation begins with a conscious command from Sullivan. If he thinks, for example, about closing his hand, the command from his brain travels in the form of a low-voltage electrical current along the nerve, to a band of muscle in his chest. The chest muscle is key for Sullivan, because it is the locale at which various nerves from his amputated hand and arm have been reattached, or re-innervated, during previous surgery. Because that portion of Sullivan's hand nerve is still intact, it is able to deliver the "close hand" command to the chest muscle. The chest muscle therefore contracts, and a myoelectric sensor atop Sullivan's skin detects the contraction (muscle contractions naturally emit an electric field), sends it to an amplifier and then to a digital signal processor (DSP) located in the Boston Digital Arm. The arm's DSP interprets the signal, then sends a command to the hand motor, which closes the hand.
All of this happens in the blink of an eye. For engineers, the challenge is to keep the signals moving at a rate that's roughly comparable to the natural rate of travel through the human body, so there isn't a noticeable lag time between Sullivan's thoughts and his arm movements. To do that, the signals are sent from the myoelectric sensors to a 1.75-inch diameter interconnect board in the arm and through to a differenetial amplifier, before reaching the DSP. The DSP, operating at 20 MHz, sorts through one or more signals at a time, then sends commands to the appropriate motors.
At the end of all this activity, the arm's response must not only be quick, but accurate as well. If Sullivan wants to move in a powerful fashion, the bionic arm must respond in kind.
"It's got to be linear," notes Richard Weir, a research scientist at the Jesse Brown VA Medical Center, professor at Northwestern University's Biomedical Engineering Department, and designer of the humeral rotator. "Contract the muscle hard; create a bigger signal; drive the motor faster."
Interpreting Neural Signals
Understanding the body's signals, however, may be the most daunting task of all for engineers. For that, the DSP must be fast and capable, and the on-board software must be able to determine which motor, what direction it should turn, and how fast.
To deal with the hardware challenges, Liberating Technologies employed a Texas Instruments (TI) C2000 DSP, rather than a conventional microcontroller, because DSP provides greater number-crunching capabilities. "DSP gives more advantages in this case, mainly in the number of motors it can control, as well as in packaging and integration," says Chris Clearman, C2000 business development manager for TI.
Making the decisions, however, was a matter for the software running on the C2000. Determining which motor to run was probably the simplest of those decisions, since motors have dedicated muscles assigned to them. A 2×6-inch printed circuit board inside the Boston Digital Arm incorporates six motor drives, each connected back to a separate band of muscle. For example, one band of chest muscle might command wrist pronation, while another commands hand grasp.
The task of deciding which direction and how fast to run the motors, however, is a stickier issue. To do that, engineers at the University of New Brunswick worked with Kuiken to implement pattern recognition algorithms that "look" at the incoming signals from the electromyogram (EMG) sensors atop the skin.
"The EMG signals are much more random than what you typically deal with in engineering design," notes Englehart of the University of New Brunswick. "It makes measurement very difficult."
Still, engineers have found ways to reliably interpret the signals. Englehart's software does this by looking at the signal patterns produced by Sullivan's muscles during prescribed tasks, and then learns from them.
"When he tries to elicit control of an elbow, wrist, or hand, the system looks at the muscle activity, and then uses neural networks to learn from those motions," Englehart says.
Based on those patterns—and based on the ability to discriminate between legitimate activity and electrical noise—the system primarily uses signal amplitude to determine motor force. It also determines which of 25 arm actions are desired within an accuracy of about 96 percent, Englehart says. Eventually, researchers hope to draw more information from various signal characteristics (particularly frequency), and then use them to map out more complex arm motions.
"There's a ton of data in there and we're still finding ways to sort it all out," Kuiken says. "Eventually, we hope to use it to determine what kind of hand grip Jesse might want to do, whether it's a fine pinch or a power grip or a key grasp."
Enhancing the Concept
For now, Sullivan's arm offers the advantage of simplicity. Sullivan doesn't need to press a button to operate the device, nor squeeze a near-by muscle that would ordinarily be unrelated to the task at hand.
"This arm offers two advantages," Kuiken says. "It gives more control information and it's more intuitive for the user."
For now, Kuiken plans to add more capabilities to Jesse Sullivan's prosthesis. He wants to enhance Sullivan's sense of touch and wants to provide dexterity to his fingers. He says he has noticed that the skin around the reinneverated areas of his chest muscle have developed sensation, and he would like to use that to provide more feel in the hand and arm. Moreover, he believes that in the future he can add more degrees of freedom to the hand and eventually create fingers that wiggle.
"We are definitely moving toward being able to control digits," he says.
"It's the same as Luke Skywalker, except that Jesse does not have the dexterity," adds Hanson. "But it will happen some day, and it won't be too far in the future."