A veteran of the Iraq war yearns to perform normal activities after losing a hand to a roadside bomb. A stroke victim wants to regain use of a partially paralyzed limb. A surgeon searches for a safe method to operate on a beating heart.
Design engineers, working with medical professionals, are devising solutions to all these challenges, in large part by implementing cutting-edge motion systems. Engineers at Johns Hopkins University, for example, are leading a global effort to design the most sophisticated bionic arm ever.
In Massachusetts, two young MIT engineering grads have devised a motorized brace to re-educate muscles in stroke victims. And at Carnegie Mellon's Robotics Institute in Pittsburgh, an engineering professor has developed a micro-robot that can literally crawl across a heart to perform medical procedures.
All these systems, as the following case histories show, pioneer advanced motion control, actuation and sensing technologies that promise to help patients, as well as spawn new ideas for engineers in other applications.
Bionic Miracle: Revolutionizing Prosthetics 2009
At Johns Hopkins University's Applied Physics Lab (APL) in Laurel, MD, Stuart Harshbarger and his team are leading a project that most engineers would consider impossible: Design a bionic limb that closely mimics the look, feel and movement of a human arm and hand.
Backed by $40 million in funding from the Defense Advanced Research Projects Agency (DARPA), APL and some 30 partners from academia, government and industry are attempting to develop this miracle limb by the end of 2009. The obvious beneficiaries: Hundreds of military veterans who have undergone amputations since the beginning of the Iraq War in 2003. In fact, one of the engineers working on the project — Jonathan Kuniholm — lost his right arm below the elbow while a Marine in Iraq.
Launched in 2005, this “Revolutionizing Prosthetics 2009” program, in its first 18 months, has developed two generations of prototypes, explored a host of alternative actuation and control schemes and established a virtual design environment for its global design team.
In August, the team unveiled two different Prototype II designs at a DARPA technical conference. Each featured 25 degrees of freedom and exhibited strength and movement speed nearly rivaling that of the human arm. Compare that performance to the team's Prototype I, whose eight degrees of freedom still outperformed all previous prosthetic designs, including an advanced bionic arm from Liberating Technologies of Massachusetts.
“This is a much more polished version of the 6-motor arm that we put together with Liberating Technologies,” says Richard Weir, a research scientist and professor in Northwestern University's Biomedical Engineering Dept. “And the electronics are much more sophisticated, relying on a bus architecture, rather than a central controller in that initial design.”
Adds Harshbarger, “The primary difference in the limb we are ultimately attempting to design is that it will be fully self-contained, with all components and power sources enclosed within it. Yet the weight must be comparable to that of the human arm — about seven pounds.”
A Look at Prototype II
As it moves toward its final design over the next two years, the team will select the best options from technologies already unveiled in prototypes and will also introduce new concepts that have not yet been integrated in the early prototypes.
Prototype II designs included an “extrinsic design,” where all of the actuation is in the forearm, much like the native anatomy, and an “intrinsic hand,” in which most of the actuation components reside in the hand and wrist.
The “intrinsic hand” in Prototype II has 18 degrees of freedom within the hand and wrist, each of which is actuated by customized 8W brushless micro-motors. The upper arm in the design has four more motor-actuated degrees of freedom.
“Getting all the force, torque and range of motion that DARPA desires in an artificial limb that is the same size and weight of a human arm is a huge mechanical challenge,” says Tom Van Doren of New World Assoc., Fredericksburg, VA. “We had to focus on such things as designing a motor controller that actually fits inside a prosthetic finger.”
The engineers met that challenge with a dime-sized flexible circuit that wraps around the finger and can handle up to four amps. The board also houses a microcontroller that runs all the commutations for a brushless motor, as well as sensors for position and force.
Within three of the fingers, there are three motors, three controllers, as well as tiny transmission components like clutches and ballscrew drives. On knuckle joints, you'll find a motor connected to a rotor clutch, belt drive and a high-ratio drive.
“It's like a watchmaker's art,” says Van Doren, who gives credit for much of this miniature mechanical design to the Austria-based Otto Bock design firm.
The alternate Prototype II design — the “extrinsic hand” — features a “cobot” developed by Kinea Design. The cylindrical device is located in the forearm and contains most of the key prosthetic components. A single customized 40W brushless motor in the cobot runs a 15-output, continuously variable transmission that pulls on 14 “tendons,” made of Honeywell's high-strength, Teflon-coated Spectra fiber. Three of the tendons are routed to the wrist and the rest to the hand. There is also one rotary output for wrist rotation.
“On the positive side, this design delivers a mechanical impedance that is closer to what we have in a human hand,” says Van Doren. “The negative is that the tendons are more prone to breakage.”
Adds Northwestern's Weir, “The extrinsic design also presents a tougher challenge because you need to control multiple CVTs to get coordinated motion.”
The electronics for both design concepts also are complex. In the early prototypes, central motor control electronics, which reside in the upper arm area of the prosthesis, provide motor control and communicate with position feedback sensors in joints, which must be able to handle 26 amps for short periods.
In the Prototype II designs, there are about 25 microprocessors, 80 tactile sensors integrated into the finger tips and hand, and additional positioning sensors for the finger joints and larger joints of the limb. And this is just the foundation of even more sophisticated sensory feedback capability in the final design.
Michael Bridges, a control theory expert at APL, describes the limb's control structure as “hierarchical” in nature. “There are nested levels of control loops,” he says. “At the lowest level, the finger, you have individual control loops, but higher up you have a centralized hand controller that must send commands to each finger to get a coordinated grasp. And if you go to the upper arm, you may want to control the velocity of the wrist moving in a specific direction. And that requires coordinating four different motors in the upper arm.”
Van Doren says these prototype architectures represent the biggest technical challenges the project's engineers have ever faced. The intrinsic hand, for example, contains just under a 1,000 parts, and the cobot design about 600.
Linking the Design Team
While designing Prototype I, the team also laid the groundwork for a virtual design environment. Engineers designed a wireless interface in the prosthesis that communicates with this virtual integration system. Researchers can move the prototype limb, for example, and that same motion appears on the computer screen.
“Within this virtual environment are building blocks for signal processing algorithms, as well as control system algorithms, so we can explore different control strategies. It also allows us to evaluate the mechanical properties and kinematics of the limb,” says Harshbarger.
In short, this virtual integration, based on MATLAB, Simulink and proprietary software, allows APL to synchronize the work of its partners in distant locations. “This is a highly synergistic effort where everyone is providing expertise that is directed toward meeting the overriding goals of the program,” says Harshbarger.
The virtual environment also gives patients the opportunity to try the limb and give their feedback before any prosthetic is manufactured. Kuniholm, the Marine veteran, has worked with the limb's pattern-recognition software to interpret his commands and turn them into motions that appear on the computer screen. In just 45 minutes after putting on Prototype II, he was controlling it with no problems.
“Depending on the patient, there could be a number of different user inputs to drive this arm,” explains APL's Bridges. “With each of these, there could be a wide range of underlying control techniques and this virtual integration environment allows for a very flexible and modular software test environment.”
Pioneering Future Technologies
Much remains to be accomplished before the deadline for final design in late 2009 and the prototypes developed so far are by no means the last word. Researchers have been working on many unusual actuation and control schemes. As an alternative to electromechanical actuators, for instance, Michael Goldfarb of Vanderbilt University is exploring a system that would actuate pneumatic cylinders with steam from hydrogen peroxide reacting with an iridium catalyst. Another hydraulic version would use a dc motor to drive a pump and would store up energy in an accumulator.
Engineers also need to develop more advanced control and sensory feedback mechanisms. The myoelectric control design for Prototypes I and II relies on surface electrodes. In a procedure developed by Dr. Todd Kuiken of the Rehabilitation Institute of Chicago (RIC), a surgeon first reroutes nerves from the arm stump of the amputee to the chest area. When the amputee contracts these chest muscles, surface electrodes sense this activity and convert it to a control signal for the prosthesis.
The problem, according to Northwestern's Weir, is that these surface electrodes can only provide up to four channels for controlling the prosthesis because the signals are prone to cross-talk. To pick up more signals and thereby generate a much broader range of motions in the prosthesis, Weir is working with other researchers, including Phillip Troyk of Illinois Institute of Technology, on a system that would feature implantable myoelectric sensors (IMES). These tiny hermetically sealed devices, about the size of a large grain of rice, would be injected into muscles in a grid fashion to achieve more stable and robust signals. The sensors receive their power and command signals via a transmitter coil outside the body.
The engineers also are researching other sensory feedback and neural interface mechanisms to give prosthesis users a sense of touch, temperature, pressure and vibration, as well as a sense of where the limb is situated in space. Currently, the principal method for achieving this feedback is the use of a “tactor,” an electromechanical device situated on the surface of the chest. New options may include a device, about 5 mm sq, which will feature 100 electrodes, each with its own amplifier and circuitry and a wireless interface.
Then there's the matter of weight. The prototype II designs each weighed around 9.5 lb — about a third more than the limit — and that includes an exterior battery and control electronics. These components must be downsized and enclosed within the prosthesis in the final design.
But these and other tough challenges haven't dampened the excitement surrounding Revolutionizing Prosthetics 2009. “I've never seen a group of engineers so motivated,” says Van Doren.
HeartLander: New Option for Minimally Invasive Surgery
What started for Cameron Riviere as a project to develop a handheld instrument to mitigate hand tremor during microsurgery on the eye took a very curious turn. The Carnegie Mellon University professor ended up inventing a miniscule robot that can actually navigate the heart.
As the researcher at CMU's Robotics Institute tells it, he got sidetracked with the challenge of helping heart surgeons in their quest to find ways to operate on a beating heart, rather than stopping the heart and placing the patient on a heart-lung machine. “I began to wonder whether any of the techniques I had developed for tremor could be applied to heartbeat motion.”
After observing surgical procedures and holding discussions with his medical collaborator, heart surgeon Marco Zenati of the University of Pittsburgh, Riviere eventually concluded that developing a miniature device that could be placed on the heart and controlled by a surgeon would be the best means of operating on a beating heart, where the amplitude of motion can range up to 2 cm during beats.
“Given how large that heart motion is and how small an area you have to work in, it struck me that an approach like this would be a simpler and more economical solution than designing an exterior robotic device that would need to oscillate to match the heartbeat motions,” he says.
His solution, which has already been tested in open-heart surgery on pigs, is called HeartLander. Measuring 5 mm tall, 8 mm wide and 2 mm long when contracted, the device literally crawls across the surface of the heart in inch-worm fashion. Depending on the medical procedure, the tiny robot can be equipped to burn away diseased tissue, inject drugs or stimulate the heart muscle with electrodes.
The surgeon inserts HeartLander into the body through an incision directly below the sternum. After making another incision in the pericardium, the sac that encloses the heart, the surgeon places the robot directly on the surface of the heart.
“You're letting the device move freely on the heart and are solving the problem of motion compensation in a much easier, passive way versus a large, expensive active robotic system that would need to operate in multiple degrees of freedom,” says Riviere.
Controlling the Motion
For locomotion, HeartLander relies on two suction cups embedded in a plastic housing — one in the front of the device and one in the rear section. Vacuum pressure from an external pump is delivered through two lines contained within a flexible tether that runs through the device. By alternating the suction from the front and rear body sections, the system changes the distance between them and moves the device in 10 to 15 mm “steps” to the desired location on the heart. To “steer” HeartLander, the tether also contains two drive wires, which pass through the rear section of the device and connect to the left and right sides of the front foot of the device. The drive wires are actuated by two exterior stepper motors, each with a running torque of 0.8N.
The surgeon controls HeartLander's location, as well as the needed therapy, using a joystick. A graphical interface, constructed with MATLAB software, shows the exact location of the robot on the heart. Real-time location is measured using a miniature 6-DOF magnetic tracking sensor, Ascension Technology's microBIRD, which is located in the back section of the device.
Riviere says much work must still be done on the system before it is ready for clinical trials. Besides further animal studies, researchers need to develop various end effectors for HeartLander, as well as establish interfaces between the HeartLander system and medical imaging devices. Even so, the CMU professor envisions the device being used someday in a number of different therapies, including myocardial injections, placement of electrodes for pacemakers or as a carrier for fiber-optic probes for laser ablation.
“We really see the device as a delivery vehicle that gets you on the surface of the heart for a whole variety of different therapies. And it allows you to operate on a beating heart, which is better for the patient,” Riviere says.
Myomo e100: Restoring Body Movement
In 50 percent of the cases, stroke victims suffer one-sided paralysis that often dictates many months of rehabilitation. Yet only about 5 percent of such patients regain full arm function.
While doing engineering graduate school work at MIT in robotics, John McBean and Kailas Narendran began a project that promises to reverse that gloomy prognosis. Working under MIT design professor Woody Flowers, the two received a foundation grant to work on a new therapeutic device. Five years later, after a series of prototypes and clinical studies, their work has resulted in an FDA-approved medical device for stroke rehabilitation called the e100 NeuroRobotic System.
Worn as an arm brace, the device harnesses exterior surface electromyography (EMG) sensors, placed on the biceps or triceps, to detect and monitor a person's electrical muscle activity. System software filters and processes these signals and sends the data to the wearable robotic device, which provides assistance in the desired movement.
“We used off-the-shelf EMG sensors, which are basically thumb-sized sticks that do minimal signal conditioning and pre-amplification,” says Narendran, whose specialty is electrical engineering and computer science. After signal preconditioning, a PIC microcontroller provides digital signal processing.
Software for the initial prototype was in C++ and the engineers did initial analysis of the system in MATLAB. Later on, they used Octave, an open source software package, for simulation and analysis. System code for the commercial device is in C.
The wearable parts of the system weigh less than 1.5 lb and include a 10W dc motor and miniature gear train located in a housing behind the elbow. These drive components apply lifting torque to lightweight cups that cradle the arm. A 40W nickel metal hydride battery, residing in a back pack, powers the system.
“The system has to supply enough force and torque to help lift the arm,” says McBean, “but it can't be too heavy or uncomfortable for an already compromised limb.”
In contrast to rehabilitation equipment that stroke survivors might use in a gym setting, the e100 is designed as an aid to retrain neurological pathways for everyday activities. “The goal was to create a closed-loop system for learning,” says McBean. “The user's brain remains in the feedback loop. The device takes its cue from the user and provides assistance and the user perceives the resulting movement as a reward and is encouraged to try again.”
Clinical studies in 2005 with stroke survivors at Boston's Spaulding Rehabilitation Hospital, in conjunction with Dr. Joel Stein, yielded some very encouraging results. The patients used the device for therapy for just 18 hours over the course of six to nine weeks. On average, patients experienced a 25 percent reduction in arm tightness and a 23 percent reduction in arm impairment.
“This was very compelling, given that patients used the device for a very limited period,” says Narendran.
A new company called Myomo has been formed to market the device, slated for launch in the first quarter of 2008. Costing $5,000 to $10,000 and available through rehabilitation facilities, the e100 could find plenty of potential users. More than 700,000 new stroke cases occur each year in the U.S. alone. The device's inventors also see the technology evolving into a possible family of rehabilitation devices.