Consider this: how intuitive is it for you to reach out and grab the nearest object to you? Perhaps you are reading this on a mobile device that you are holding in one hand while doing something else entirely with the other one. Have you ever contemplated how many mundane tasks you perform on a daily basis that require two hands? What about tasks that require fine motor control, such as tying your shoelaces or handling something delicate, like an egg? Our world is filled with life-changing technology and recent decades have yielded more astonishing advances than most of us could ever have imagined, but there is one area of critical importance to a huge number of people that is only now beginning to catch up. Prosthetic devices have been around since the early 1800s, but the designs of such devices have tended to go through long periods with little or no significant innovation. During the latter half of the 20th century, many prosthetic devices underwent only incremental design improvements. Since the turn of the millennium however, there has been dramatic progress, especially in the design of upper-limb prostheses. As a result of these advances, upper-limb prostheses are becoming more user friendly and, in the case of some cutting-edge devices, are now much closer to being comparable with the abilities of a biological arm. Unfortunately, the cost of such advanced devices places them beyond the reach of many users. Even for those who are fortunate enough to be able to afford one, practical issues can arise at the interface between the residual limb and the device. Where signals recorded from the residual limb are used to control the prosthesis, the quality of signals that can be obtained (which varies greatly from user to user) is of paramount importance.

Prosthetics and assistive technology

[caption id="attachment_33248" align="alignright" width="169"]prosthetic-limb upper-limb-prosthesis[/caption] My interest in prosthetics began in 2011 during my Leaving Certificate Technology course. As part of the prescribed project in that subject, I delved deep into the world of assistive technology and its applications. In September 2015, during the final year of my electrical-engineering degree in the Dublin Institute of Technology, I approached Dr Ted Burke, lecturer in DIT’s School of Eletrical Engineering Systems. I proposed revisiting the topic with the aim of developing a novel method by which a signal (or signals) recorded from a user’s forearm could be used to easily operate a prosthesis and control the tightness of the end effector (hand). The starting point for my investigation was to research the mechanics of the human arm, the biopotentials (electrical signals) that can be recorded from various muscles in the arm and which of these signals might provide a viable means of controlling a prosthesis. Consulting academics with a background in this field as well as clinicians in IDS Cappagh (whose support was an invaluable asset in this study) greatly sped up my learning. For those unfamiliar with the physiology of voluntary muscle control, consider this example: when we decide to move or activate a muscle, impulses are emitted by a motor area in the brain. These impulses propagate at high speed along the nerve fibres of the peripheral nervous system to the relevant muscle, where they trigger a series of chemical reactions that result in the physical contraction of muscle cells. The activation of each muscle cell involves a tiny movement of electrical charge across the cell’s membrane. The result of many muscle cells activating together is a biopotential signal called the electromyogram (EMG) that can be measured in the form of a time-varying voltage on the surface of the skin. The EMG is a noisy random signal, with an amplitude that is typically on the millivolt scale. The more intense the muscle contraction, the greater the magnitude of the EMG. Research into industry standards as well as the anatomy of residual limbs pointed to the flexor carpi radialis (herein referred to as the anterior of the forearm) and the flexor carpi ulnaris (herein referred to as the posterior of the forearm) as the most suitable muscles from which to measure control signals. To locate the anterior forearm muscle, place your right hand on the posterior (outside) of your left forearm near the elbow with the palm facing down, then flex your left wrist up and down to activate the muscle. To locate the posterior forearm muscle, place your right hand on the anterior (inside) of the forearm with the palm facing up and flex your left wrist up and down again. Clenching your left hand into a fist activates the anterior and posterior simultaneously. These three distinct types of muscle activation are referred to below as anterior flexion, posterior flexion and code contraction (both anterior and posterior activated simultaneously). It was these three gestures that I used to control my prosthesis, via a simple state machine model.

Development of biopotential instrumentation

[caption id="attachment_33296" align="alignright" width="300"]state-table-flowchart CLICK TO ENLARGE: State table flowchart[/caption] Having selected a specific muscle location on which to focus the signal measurement, development of the biopotential instrumentation began. Surface-mounted electrodes are the industry standard. IDS Cappagh is affiliated with Ottobock, which uses myoelectrodes in their electromechanical arms. Myoelectrodes are small units about the size of a standard flash drive which contain surface electrodes, a differential voltage amplifier, signal rectifiers and signal conditioners. These would have been suitable for use in the project but unfortunately are quite expensive. Due to budget limitations, I opted to develop my own signal-acquisition system to be used in conjunction with surface-mounted electrodes. The first obstacle to overcome with regard to obtaining a signal was the development of suitable electrode cables. Standard biopotential electrode pads were available in the college for use with clinical measurement systems. However, matching connectors were not available. After some experimentation, the connector on the electrode pads turned out to be almost identical to a snap button that can be sewn onto clothing. After acquiring several different brands of snap connector and testing them for continuity, I manufactured cables that facilitate a good quality connection, but also allow for interchangeability and ease of integration into the amplifier circuit. Using some cable, the female end of a snap button, header pins and some hot glue to insulate the joints, I created my own electrodes. Having established a reliable method of connection, the focus turned to development of circuitry to amplify and condition the signal. The microcontroller used was a dsPIC30F4011, which requires an input signal of significantly greater amplitude than that obtained directly from the surface of the skin. Hence, amplification was essential. Furthermore, to minimise code complexity, I elected to input each muscle’s signal in binary form only (high for muscle active/low for muscle resting). These binary signals act as inputs to a state machine model that controls the movement of the prosthesis.

Designing the circuit

[caption id="attachment_33249" align="alignright" width="300"]prosthetic-limb1 upper-limb-prosthesis[/caption] With these two stipulations in mind, circuit design commenced. Both anterior and posterior signals were treated the same and both will be referred to below simply as the ‘signal’. The differential voltage signal recorded from the electrodes was initially passed through an instrumentation amplifier with a gain of 11 to produce a higher amplitude single-ended signal. From there, it was fed into a non-inverting amplifier with a gain of 10, resulting in a final signal amplitude on the volt scale as required by the dsPIC30F4011. At this point, we have an analogue signal that is viable to work with. However, it was observed that oscillations were present in the signal regardless of whether the muscle is at rest or active. This was likely the result of interference from external sources, which is a familiar problem in biopotential measurement. At times, the resulting degradation of the recorded signal was sufficient to make the state machine (and hence the prosthesis) behave unpredictably. I therefore implemented a simple envelope detector to smooth out the signal, which made it much easier for the user to reliably select the desired level of activation (which is performed using muscle flexion gestures). It was desired that this project would be easily adaptable to the wearer and the electrical biopotential signals obtainable from their arm. Even in fully able arms, significant variation in signal amplitude and quality is observed between different subjects. This gave rise to the next development, which was to devise a way in which a binary (on-off) signal could be easily utilised with the existing code while also allowing any adjustments needed to be made in the hardware. A comparator circuit with variable threshold was chosen as the most suitable augmentation to complete the measurement electronics. The threshold level of each signal’s comparator can be set using external potentiometers (dials) that are mounted externally so that the user can adjust the sensitivity to muscle activation at any time.

Gesture sequence to control the prosthesis

20160519_183436Now that the electronics could reliably measure a signal from the arm and convert it to binary form, indicating whether or not each of the two selected muscles is active or not, it was time to finalise the gesture sequence the user would employ to control the prosthesis. Many existing systems differentiate between users with strong and weak muscle signals. Those with stronger signals are granted proportional control of hand closing and gripping using signal amplitude, whereas those with weaker signals are typically limited to using the more easily detected code contraction to open and close the hand. My desire was to develop a system that would facilitate variable grip control, even for those with weaker signals, by means of pre-defined gesture patterns. Once the preset threshold level has been set to facilitate reliable detection of intentional activation of each muscle, the user controls the prosthesis using a series of anterior, posterior or code contractions (fist clenches) in a specific order to select and then activate a desired level of grip. When the prosthesis is powered up, a fanfare of lights indicates to the user that initialisation is in progress.Once initialisation is complete a yellow LED lights to indicate that the arm is in standby mode and at rest. When the user performs a code contraction, the system observes length of the activation. If it is less than three seconds in length, the system enters grip set mode (indicated by a flashing yellow LED); if the duration of the code contraction is longer than three seconds, the arm skips straight to the closing procedure. Grip set is a mode whereby the user can increment or decrement the level of tightness they wish to obtain by executing an anterior or posterior contraction. A row of red LEDs on the exterior of the arm light up to indicate the level (ranging from 0–4 or fully open to fully closed). Once the desired level has been obtained, executing another code contraction begins the tightening procedure. This is indicated by a flashing red LED and can be stopped at any time should the user need it by either performing a code contraction to return to grip set mode (should the wrong level have been selected), or an anterior flexion (to immediately open it). A green LED lights to indicate the end effector has reached the user’s desired level of grip. The arm remains in this state until the user releases it by performing a code contraction. A red LED lights to indicate that the end effector is opening. Once it has returned to the open position, the state machine returns to the aforementioned standby state. The final phase of the project was the mechanical design of the prosthesis. The end effector employs a simple design comprising multiple lengths of flexible conduit, incorporating string threaded through it to act as tendons. The end effector closely resembles a human hand with good biomimetic properties. The strings which act as tendons for each finger terminate at a single point connected to a bell crank mechanism which in turn was connected to a ratchet and pawl (operated by a stepper and servo motor), which was used to retain and release the tension in the end effector during gripping actions. This structure is simple to control, places minimal stress on the actuators and allows de-energisation during gripping and resting, which reduces power consumption.

Conclusion

The resulting implementation provides a successful proof of concept that, with further refinement, could be form the basis of an effective and useful device. As a result, this project was nominated for the 2016 Siemens Innovative Student of the Year Award, run by Engineers Ireland. My experiences working on this project have also reinforced my interest in biomedical engineering, especially in the use of instrumentation and software in medical devices and assistive technology. Having graduated with a BEngTech in Electrical and Control Systems Engineering, I am now completing a BEng in Computer and Communications Engineering in DIT with the hope of continuing my work in the field of prosthetics. More can be read about this project and others at aislinglee.wordpress.com. Aisling Lee, BEngTech is a recent graduate of DIT’s Bachelor of Engineering Technology in Electrical and Control Systems. She has returned to DIT and is currently working towards a Bachelor of Engineering in Computer and Communications (part of a scholarship she is in receipt of from the ESB). It was as part of their engineering program she served two years as an apprentice electrician/network technician. Her main academic fields of interest are robotics, user-friendly software development, electronics and their interdisciplinary uses. Most recently, she has began to develop an interest in biometrics and its application. Her final-year project was based in this field and earned her a place as one of six finalists in the Innovative Student Engineer of the Year Awards.