Losing an arm is a traumatic, life-changing event. Too many military personnel suffer that loss. Even in peacetime, as many as 20 armed services members endure upper or lower limb amputations annually due to noncombat accidents. To help wounded warriors return to active and healthy lifestyles, the Department of Veterans Affairs not only provides medical assistance, prosthetics, training, and other assistance, it also funds R&D into better prosthetics.
At the Advanced Platform Technology Center (APTC) at the VA Medical Center in Cleveland, one team of biomedical researchers is working to develop technology that would give veterans (and civilians) with upper-limb losses a sense of feeling in their prosthetics. They are installing sensors on the artificial hand portion of the prosthetic, then processing and routing those sensor signals to the user’s brain via the nerves that once served the missing hand.
The goal of the project, according to team member Assistant Professor Dustin Tyler, Director of Engineering, Quality, and Regulatory Affairs at APTC and a faculty member in nearby Case Western Reserve University’s Biomedical Engineering Dept., is determining what kind of sensations the sensors on prosthetics can provide users, how many sites can be given this artificial sense of touch, and what technologies can best do the job.
Sensors
“The sensors we currently use are thin-film, force sensor resistors (FSRs) from Tekscan Inc.,” says Tyler.
An FSR consists of two electrodes separated by a thin sheet of material. Applying pressure moves the electrodes closer together and changes the sensor’s resistance as a function of pressure. FSRs are simply taped to the artifical hand, a classic three-jaw hand prosthesis from Ottobock with one degree of freedom — the index and middle finger move together toward and away from the thumb. The center’s current prosthetic has four FSRs that measure the finger opening or span between the thumb and fingers, as well as pressure at the tip of the thumb, index finger, and middle finger.
“These let users grasp and release objects and give an indication of how much force they are exerting in their grasp. But in the future, sensors need to be more rugged to withstand daily use without breaking down. And they will be mounted so they are protected. But there are some drawbacks to ‘hardening’ sensors,” notes Tyler.
“One medical-device manufacturer puts sensors on the internal structural metal parts of its prosthetics. This provides a level of protection, but limits the range of forces they can detect,” Tyler says.
“Down the line, as we try to detect forces distributed over artificial hands, we will need to find mounting locations that let sensors detect surface sensations and pressure with better resolution, and not just a gross output of total forces but better spatial resolution,” says Tyler. “We will always want to detect feelings from fingers, so sensors will likely mount on the prosthetic’s fingertips, which means they will have to be protected against wear and liquids.”
In APTC’s current setup, an exterior battery powers the sensors. Actual prosthetics, however, would not tether users to external batteries or power supplies.
“Batteries are getting more capable, and motors and sensors are becoming more efficient, so the ideal would be to have a single-charged battery pack last all day,” says Tyler. “But until then, biomedical engineers will likely design artificial hands and arms with quick-change-out battery packs holding a minimum of 4 hours of use.”
The PC and nerve stimulator
Raw sensor data gets amplified, filtered, digitized, and then sent to the PC at no more than 100 Hz, which is enough bandwidth for an artificial hand’s sense of touch. The PC, a standard desktop model running Windows, records all the sensor data and performs some numerical processing on it using Matlab (from Mathworks).
In general, the PC maps the sensor information to stimulation patterns for nerves that will generate the desired sensations in users. Currently, only researchers can adjust or change settings on the PC and stimulator, letting them update programming and try new combinations of stimulation pulses. That’s one reason the stimulator and PC will likely remain separate even in later versions, says Tyler.
“In the future, users might be able to control some gain parameters so that the hand is more sensitive for doing delicate work and less sensitive for heavier-duty work,” explains Tyler. “But it would likely be nothing more sophisticated than a volume knob.”
The relatively simple stimulator, the product of biomedical engineers at the VA's Functional Electrical Stimulation Center (FESC) in Cleveland, creates three current-controlled pulse trains. The three stimulation signals get sent to the three major nerves that usually carry sensory signals from the hand to the brain: the median, ulnar, and radial nerves. This stimulation is ac in nature and biphasic with balanced current flow in successive negative and positive impulses. Biomedical engineers long ago discovered that long-term monophasic stimulation with all positive or all negative pulses creates chemical and charge imbalances that break down nearby blood vessels and muscle tissues.
The APTC project’s goal is to eventually replace the PC with an embedded processor that would be small enough to mount inside the prosthetic. It would pull in sensor data, do some processing, and match incoming sensor data with the proper nerve stimulation signal (pulse, timing, and amplitude).
“One long-term approach is to use prosthetic-mounted sensors that detect the raw haptic data, which will be processed in a module attached to the artificial hand, perhaps no larger than a wristwatch,” says Tyler. “This would communicate with a stimulator implanted in users like a pacemaker but send sensory signals to the proper nerves. The prosthetic would also record the muscle activity (electromyogram, or EMG) of the user trying to control his missing hand. These EMG signals would get processed and be used to control the prosthetic’s drive motors that move the digits and hand.”
The right stimulation
One of the most difficult but intriguing parts of the APTC project is determining what stimulation to apply given a specific sensor input.
“Biomedical engineers started by copying the paradigm used to control the muscles of people with spinal-cord injuries. It relies on consistent trains of pulses sent to muscles via efferent nerves (those that go from the spinal cord to muscles). The higher the pulses’ amplitude, the more muscle fibers get excited, so the muscle generates more force,” Tyler says.
“But this doesn’t work on afferent nerves (which carry information from sense organs to the brain). Users felt sensations, but they weren’t natural. Instead, they got that ‘pins and needles’ feeling, like their fingers or hand had fallen asleep. This wasn’t too useful, but to people who had never felt anything from their prosthetics, it was better than nothing,” he explains.
These results were not surprising. Sensation is much more complex than muscle movement. For example, muscles do little if any processing on incoming nerve signals, while the brain does a host of processing on incoming sensory nerve signals. That’s because sensory signals include more information, which could be encoded in the signal’s frequency, the pulse duration and amplitude, the space between pulses, or the interference or combinational effects of signals travelling along the same and nearby nerves. “Or the information might be a contained in a permutation of all these parameters,” notes Tyler.
Over time, Tyler and his team have gone far beyond letting users just feel how tightly they are grasping an object. “In fact, we’ve elicited feelings and sensations in as many as 19 different places in the users’ missing hands,” says Tyler. “This includes a couple places in the palm, the fingertips, a few places on the back of the hand and wrist, and several down the outside edge of the pinky. So our technology for sensors on the prosthetic hand is behind in terms of the number of places we can elicit sensations in users.”
“We would also like to find the right stimulation that would recreate shear, which would let users ‘feel’ when something is sliding through their grasp,” says Tyler.
“Another sensation amputees really miss is warmth, especially the warmth of a child’s or spouse’s touch,” says Tyler. “Unfortunately, that would add another type of sensor to a space-limited prosthetic. And nerves that relay heat and cold to the brain are smaller than those that carry pressure information. Smaller nerves are more difficult to locate, isolate, and stimulate, and special electrodes would have to be developed. So giving prosthetics a sense of hot and cold is not our focus.”
Connectors and electrodes
In the APTC setup, different stimulation signals are sent to 20 nerve sites accessed by three implanted electrode cuffs, each on a different nerve. There is a pair of eight-contact cuffs, and a four-contact spiral cuff wraps around one of the nerves, making 20 sites. On the eight-contact electrode, for example, each of the eight wires or leads is connected to a small piece of platinum foil in contact with the nerve. Platinum is biocompatible and provides good electrical contact with the nerve tissue.
Each of the 20 leads terminate in pins, which attach to 20 corresponding transcutaneous leads via a matched-pin connector. This connector uses a spring to connect two lead pins.
Mating pins are inserted in either end of the spring, which is wound around each pin in the opposite direction of its spiral, a time consuming process. Pulling the connected leads apart tightens the spring around the pins, much like a Chinese finger trap. The spring and pins are then covered with a silicone sleeve to isolate and insulate the electrical connection from body fluids. The benefit of this somewhat arcane connector, which will be replaced in future, refined versions of the system, is that it creates solid electrical connections and is replaceable.
These second leads run under the skin up the arm to the user’s shoulder, where they exit the user’s body, leaving 2 or 3 in. of wires terminating in a multilead jack protruding from the skin.
Tyler uses flat interface nerve electrodes (FINEs) in the haptic system. Such electrodes have a two-part housing that opens like a clamshell to let surgeons place a nerve longitudinally down its center. Closing the housing holds the nerve in place with minimal force, ensuring it is not damaged and blood flow to it is not interrupted. In future versions, the bulk of the electrode will be reduced by using high-performance polymers, such as PEEK, to give the electrode good mechanical properties, as well as more precise alignment of the more closely positioned electrical contacts. Eventually, the electrodes will include thin-film componenets containing small and precisely postioned contacts, thanks to photolithographic manufacturing, the same processes used to make ICs.
Larger nerves are much like ropes in that they are made up of smaller fibers or fassicles. Fassicles are about 0.5 mm in diameter, so the typical 10-mm-wide nerve likely has at least 16 of them going from side to side. With these more refined and precise FINEs, Tyler plans to increase the number of contacts from eight to 32, or 16 on each side of the nerve. Eventually, FINEs could each have 64 contacts.
“This would let us stimulate more areas in the brain that correlate to more places on the prosthetic hand,” says Tyler. “Or if we combined thin-film technology with FINE, we could add rows of electrodes along the nerve, stimulating more nerve sections using a wider variety of different stimulus combination to give users more sensations.”
Do Prosthetic Feet Need Feelings?
Prosthetic lower limbs, especially those for people with amputations below the knee, have been very successful at replicating the needed biomechanics to get users up and walking, even running in many cases. Yet none of them give users any sense or feeling in their artificial limbs. So do prosthetic feet need haptic technology?
“Yes, the mechanics of lower-limb prosthetics are spectacular, but they are usually being used on flat, level surfaces,” notes Assistant Professor Dustin Tyler, Director of Engineering, Quality, and Regulatory Affairs at the Advanced Platform Technology Center (APTC) at the Veterans Center in Cleveland and faculty member in nearby Case Western Reserve University’s Engineering Department.
“It becomes more challenging on uneven surfaces when the artificial foot moves unpredictably because the user cannot feel the ground. So users probably could benefit from receiving foot position data, which the sense of touch should be able to provide,” Tyler says.
Climbing steps and hills also present problems to people with prosthetic feet. For example, they have to watch carefully to know whether their entire artificial foot is on a step or just a portion, and they don’t know which portion, the toe or the heel. “Giving them the sense of pressure on the bottom of their ‘feet’ would give them better capability and confidence in climbing steps,” says Tyler.
“Plus, receiving sensation from an electromechanical device, foot or hand, lets people consider the device as part of themselves rather than just a tool hanging on the end of their limb,” says Tyler.