Chapter 6D - Atlas of Limb Prosthetics: Surgical, Prosthetic, and Rehabilitation Principles
Upper-Limb Prosthetics: Control of Limb Prostheses
Dudley S. Childress, Ph.Dá
The material that follows in large part deals with control of externally powered prostheses. Prostheses that are entirely cable actuated and body powered are dealt with in another section of the Atlas (see Chapter 6A and Chapter 6B). The various control schemes of cable-operated prostheses are considered there. Nevertheless, cable-operated systems will also be considered in this discussion because they are an important form of prosthesis control, even when electric-powered components are used. This is particularly the case for high-level unilateral and bilateral amputees, where the systems of choice often use hybrid control (cable, myoelectric, switch, or some combination of these or other methods) and hybrid power (electric and body power). Also, powered systems that emulate cable systems will play, it is believed, an important role in prosthesis control of the future. Consequently, any general discussion of control systems for arm amputees must include cable control from body movement inputs.
It is interesting that when we talk about lower-limb prostheses, we seldom talk about control. Instead, we talk more about interface loads, suspension, alignment, etc. This comes about because the lower limb must bear significant body loads, because lower-limb activity is highly repetitious and stylized (e.g., walking) and because the intact knee joint acts as the natural controller for the transtibial prosthesis, the most commonly prescribed lower-limb prosthesis. The transradial prosthesis for the upper-limb (the most common upper-limb prosthesis) is similar to the transtibial prosthesis in that it is an extension of the limb and because position and velocity are controlled by the elbow joint. However, with the transradial prosthesis, the attached prehensor needs to be controlled (unless a passive hand is used), whereas the artificial foot is a passive mechanism. Hence, discussion is more prevalent about control of upper-limb prostheses than it is with lower-limb prostheses; only for persons with amputations at the knee or higher do issues of control become apparent with lower-limb prostheses.
PLACEMENT OF PROSTHESIS CONTROL IN THE GENERAL CONTEXT OF CONTROL THEORY AND PRACTICE
Control theory is a common topic in engineering and is even a topic of mathematics. However, little will be found in engineering books regarding the control approaches that are currently used with limb prostheses and discussed here. This section attempts to place prosthetics control in the wider context of general control applications. Briefly, control theory is a part of general systems theory-the part that has to do with how one goes about creating inputs to a mechanism or system in order to produce specified outputs or responses. As applied to prosthetics, control concerns how to create inputs that will cause an artificial limb to behave in a desired way. If the inputs are generated independently of outputs, we call it "open-loop" control. If the input is, at least partially, a function of the systems output variables, as is the case with most control systems, we call it "closed-loop" or "feedback" control. Closed-loop control allows a system to adjust the inputs as the system outputs are changed by external disturbances or as the operator wants to change the output to a desired value. With arm prosthesis systems, the primary feedback method is visual feedback of output position to the prosthesis user, who is the input decision maker. This concept is shown in Fig 6D-1. for both electric-powered and body-powered prosthetics systems. A more complete view of feedback relationships in prostheses has been described by Childress.
If we regard an upper-limb artificial arm as a machine that helps someone manipulate his environment, then we can consider a human-prosthesis system as a human-machine system. Human-machine systems (e.g., airplanes, automobiles, spacecraft, and other human-operated systems) have been extensively studied in the field of "human factors engineering," and many of the ideas of that field relate, at least partially, to human-prosthesis systems. Sheridan and Ferrell have written definitively on this topic, and their book deals with many of the issues of human-machine systems, particularly from an engineering viewpoint. One aspect of prosthesis control that makes it unique when compared with typical human-machine systems is the modality of human control. While in almost all human-machine systems the operator interacts with the machine with the hands or feet, this is not the case with most human-prosthesis systems. Therefore, prosthesis systems are a subset of human-machine systems that may be classified as having "nonmanual control" modalities, as opposed to the so-called manual control systems.
The class of complex human-machine systems used in industry and elsewhere, that most nearly resemble complex human-artificial arm systems are master-slave manipulators (teleoperators). With these systems the human operator remotely controls manipulators that, for example, handle radioactive materials or that work in hostile conditions like those in outer space. Manipulator arms are somewhat similar to artificial arms and hands. A main difference is that the controls of the manipulator are activated through movement and forces of the operator's hands, arms, and/or feet. Additional differences come from the engineering constraints that prostheses and manipulators are designed under. Prostheses, because they must be carried about with the user, must be light in weight; restricted in size, shape, and appearance (somewhat like a human hand/arm); energy efficient so that they can operate all day on a relatively small battery; and quiet in operation. Manipulator design is usually not constrained nearly as much by power, weight, shape, noise, or appearance requirements. Consequently, solutions to manipulator problems often do not solve prosthesis problems. However, the ways in which manipulators are designed to provide force and sensory and proprioceptive feedback to the operator in order to improve human-manipulator interaction are highly desirable in prosthetics, and these concepts should not be ignored, even though they may not be applicable directly. Conversely, solutions to prosthetics problems may have manipulator applications. Murphy points out that bilateral users of cable-actuated prosthetic arms appear to be able to perform many tasks considerably quicker than what is typical with manipulators. He attributes this to the basic design philosophy of arm prostheses.
Another similarity of manipulators with prostheses is that the first master-slave manipulators that were designed were entirely cable controlled, just as most arm prostheses have been cable controlled. Direct cable control provides good proprioceptive and force feedback in manipulators, as it does in prostheses. As manipulators increasingly incorporated power into their designs, attempts have been made to mimic the characteristics of the previously used cable systems in the powered systems. This did not happen in prosthetics when power became available for prosthesis design, but the trend may now be in that direction. Teleoperators that provide proportional force and position feedback to the site of control are often called "telechirs." As tele-operator technology advances to more remote applications, such as in outer space or under the sea, new control advances will be necessary if the operator is to have the advantages of "automatic assistance" and "feel" to help with control of the manipulator. Some of these advances may be useful in limb prosthetics.
Solutions to problems in robotics seldom have an impact on prosthesis design, partially for the same reasons that manipulator designs have not had much impact. However, manipulators are at least human-machine systems. Robots are usually under the supervision of digital computers and so are less similar to human-prosthesis systems than manipulator systems. Consequently, even though knowledge of manipulator and robot design is surely of assistance to designers of humanprosthesis systems (artificial arms), not many ideas can be translated directly between the fields without considerable modification.
SOME COMMONLY EXPERIENCED CONTROL SYSTEMS THAT RELATE TO PROSTHESIS CONTROL
The use of powered mechanisms is a common experience of everyday life, and the control systems used in these devices are often similar to those used in powered prostheses. Practical control systems for artificial limbs are, by and large, rather simple systems. Some are so simple that when we experience them in daily life we often do not identify them as control systems.
Hand-operated bicycle brake systems are familiar cable-operated control systems that are similar to the cable control mechanisms of body-powered prostheses, except that the prosthesis systems are not operated with the hands. In the bicycle brake example, applying the brakes (gripping the rim of the wheel) is analogous to pulling the cable of a voluntary-closing prehension mechanism to grip an object. The Bowden cable was invented in 1885 by Bowden, the founder of the Raleigh bicycle company, and it is probably not by chance that the Wright brothers, the builders of the first airplane, owned a bicycle shop. Cable controls have been used extensively in the bicycle and aircraft industry and also in the smaller field of limb prosthetics.
Electric-powered automobile windows that are switch controlled for powered lowering or raising are an example of a commonly experienced system that is very similar to a switch-controlled electric-powered prosthetic joint. Pushing the control switch down causes the window to be lowered. Pushing it the other way causes the window to elevate. The window will stop whenever the switch is released. Consequently, the human operator is the feedback link for positioning the window. By operating the switch and by watching the window as it moves, the operator can position the window in almost any desired vertical position. The up-and-down operation of an electric-powered projection screen is another example of this kind of control. It is "on-off" switch control where the switch is often mechanical in construction, but which could be electronic and operated in a multitude of ways ranging from capacitive touch to breaking a light beam of a photodiode. On-off control is a widely used approach to the control of prostheses, with the control ranging from mechanical switches to electronic switches operated by myoelectric signals. It provides a kind of "velocity control" where position depends upon the time of activation of the switch and the velocity of the output (e.g., prosthetic joint or car window). It should be intuitively obvious that if a car's window moves very fast it would be difficult to position the window accurately with this kind of control. Hence, effective positioning of an output such as a powered window is feasible for a human operator using "on-off control only if the velocity of the output is low enough to be commensurate with this control mode and with the limitations of the human operator. The same is true in prosthetic systems that use "on-off" control.
Lighting systems frequently use "on-off" control. Some lights have a proportional controller so that the position of a dial determines the level of light intensity. In proportional control, the output intensity is proportional to an input setting. For example, in a lighting system, intensity may be proportional to the position of a rotary resistor (transducer) that transduces rotational position into a signal that electronically sets the light intensity. This is a kind of position control input. In another kind of lighting system, the intensity may be set in the same way an electric car window is run up and down. Pushing a switch one way causes the light intensity to go up; pushing it the opposite direction causes the intensity to go down. This allows a graded response in intensity, but it is not proportional control. Since intensity is related to the time the switch is activated, this is similar to the "velocity control" already described. In prosthetics, this method is sometimes called proportional-time control because the intensity is related to the time the switch has been activated, but it should be noted that this is not conventional use for the term "proportional."
Powered drills, powered screwdrivers, and other portable powered tools are about as close to simple powered prosthesis systems and components (e.g., electric hands, elbows, etc.) as any systems that we commonly experience in our daily lives. They are self-contained and portable, contain rechargeable batteries, use dc motors, produce rotational velocity and torque, are reversible, have interchangeable end components, and have their own control systems. Drills or screwdrivers with inexpensive control systems may use "on-off" switch control. More sophisticated devices may have proportional control in which the velocity of output rotation is proportional to position or pressure at the input. In addition, some of these devices have control mechanisms that automatically try to keep the output velocity constant for a given input setting, even when external loading is increased or decreased at the output. This is an automatic control adjustment that occurs without the knowledge of the operator but that helps with accurate control of the device.
Automobile powered steering is a kind of "boosted" power system in which the mechanism of control is similar to the nonpowered case. This is a position control system (for a stationary car) in which the position of the front wheels is directly related to the position of the steering wheel. Velocity of turning of the wheels is directly related to steering wheel velocity, and forces on the wheels are reflected into the steering column. The powered system works in the same way as its nonpow-ered equivalent, except that with powered steering the required forces (torques) and excursions can be set to appropriately match the physical capabilities of a wide range of drivers. The ideas behind powered steering appear to have considerable application in control of upper-limb prostheses, where a similar kind of "boosted" power, used in combination with cable control, enables cable force and excursion to be matched to the physical abilities of the amputee using the system. Such a system maintains the proprioceptive qualities of cable-actuated systems while also providing the benefits of powered components. The author has called this "powered cable steering." In this control approach, a cable is used to "steer" a powered prosthesis joint through use of a position control system. This approach is closely related to the concepts of "extended physiologic proprioception" as proposed by D.C. Simpsonfor the control of powered prostheses.
Aircraft flight control systems for the control of wing and tail surfaces have taken a pathway of development that is similar to those taken with manipulators and automotive steering. Airplane flight surface controls have traditionally been body powered through cables. In fact, the development of cable-operated arm prostheses after World War II was considerably influenced by this cable technology through aircraft companies (e.g., Northrup Corp.) and by aeronautical engineers. Cable-actuated systems give pilots a good "feel" for the plane just as cable-operated prostheses provide "feel" for the prosthesis. The larger, faster planes that were developed after World War II often had "boosted" power for their cable controls. As noted already, in the discussion about manipulators and automobile powered steering, new prosthesis controllers may follow this same trend. More recent advanced aircraft systems, the so-called "fly-by-wire" systems, connect the pilot to the control surfaces through electrical wire connections. Nevertheless, an effort has been made to continue to give the pilot "feel" in the control stick.
Home heating and cooling systems are in our common experience. They are a class of control systems that are called regulators and attempt to keep some variable constant (e.g., inside temperature) in the face of external changes, for example, outside temperature fluctuations. This kind of controller is automatic; however, it is designed to maintain a fixed state that is set by a constant input. Regulator-type control is not generally used in limb prosthetics. On the other hand, position servomechanisms are designed so that the output tracks or follows a time-varying input. Such systems are designed so that the output position responds quickly to input position changes. The Steeper hand position controller is an example of this kind of system as applied to prosthetics. A position of the body is sensed and translated into a position of hand opening. Control systems of this kind are not too common in everyday experience. The control system that orients a powered television antenna on top of a house by rotating the antenna until it matches a desired direction that has been set on a direction indicator box inside the house is one example that comes to mind. The fact that the antenna and the direction indicator box are only linked by an electrical position indicator means that the direction indicator on the control box can easily be moved to a new direction (there is no mechanical connection to the antenna) without a sense of "feel" of the antenna's actual position at the input. The error between a new position of the direction indicator and the actual position of the antenna is used to drive the antenna's motor to reposition the antenna on the roof. Consequently, significant differences may exist between the input position indicator and the antenna while the antenna is powered to a new position. This is in contrast with automotive power steering, already discussed, where the position of the front wheels is mechanically linked to the steering wheel so that error between the steering wheel and the front wheels is always minimal and so that a "feel" for the position of the wheels is provided through the steering column.
DESIRABLE ATTRIBUTES OF PROSTHESIS CONTROL (GOALS FOR LIMB PROSTHESIS CONTROL)
There are several highly desirable attributes of control systems for limb prostheses. Some of these attributes may be difficult, if not impossible to achieve in practice. Nevertheless, they need to be stated as goals in order to stimulate continued control system improvement and development. Systems that fall short of these goals may be serviceable and practical, but we will know that they can be improved and made better. Some of the important attributes are as follows:
- Low mental loading or subconscious control. This means that the prosthesis can be used without undue mental involvement. Successful control systems enable the users to use their artificial limbs almost subconsciously, the way people commonly use their limbs. In other words, the prosthesis should serve the user; the user should not be a servant to the prosthesis. The user should be able to think about other things, even while using the prosthesis. This kind of control may require proprioceptive and sensory feedback of the right modality in order to be achieved.
- User friendly or simple to learn to use. This feature is closely related to feature 1. It means that learning to control the prosthesis should be intuitive and natural. If this is true, the user should be able to learn to use the prosthesis quickly and easily.
- Independence in multifunctional control. Control of any function should be able to be executed without activating or interfering with the other control systems of a multifunctional prosthesis. For example, a person with prostheses on both arms should be able to use each limb independently. Operation of a function of one prosthesis should not cause any activity of the prosthesis on the opposite side. A common example where independent action is not achieved is in typical cable-operated, body-powered transhumeral prostheses with a voluntary-opening hook. If the user attempts to lift a heavy load, the hook tends to open during the lifting.
- Simultaneous, coordinated control of multiple functions. This is the ability to coordinate multiple functions simultaneously in effective and meaningful ways and, of course, without excessive mental effort (attribute 1). It also implies attribute 3 in that it allows independent control of any function or any combination of functions.
- Direct access and instantaneous response. All functions, if possible, should be directly accessible to the user and without time delay. Prosthetic systems should respond immediately to inputs, if possible.
- No sacrifice of human functional ability. The control system should not encumber any natural movement that an amputee can apply to useful purposes. In general, it is not wise to sacrifice a useful body action for the control of a prosthesis. The prosthesis should be used to supplement not subtract from available function.
- Natural appearance. If possible, the control system should be operated in ways that have a nice aesthetic appearance. Likewise, the mechanical response should be graceful, if possible. Control methods that allow aesthetically pleasing action (e.g., smooth, flowing, graceful movement) are important to prosthetic appearance, just as are shape and color. Movements that appear mechanical in nature may not be pleasing to the eye.
WHAT SHOULD BE DESIGNED/WHAT SHOULD BE CONTROLLED?
The question of what should be controlled by prosthetics control systems depends importantly upon the philosophy of artificial limb design. If the objective is to design an artificial arm that emulates a human arm as much as possible, then one may want to control joint compliance and other variables that may make the prosthetic limb have a number of characteristics of a human arm. However, at this stage of arm prosthesis development these concepts are still not clinically available. In any case, it is difficult to "replace" the human arm and/or hand system and probably always will be. Hence, the clinical approach most often taken with arm prostheses is to design them as "tools" that the amputee can effectively use in assistance with activities of daily living. That is the direction taken by the author in this chapter. It is, of course, desirable for the "tool" to look and function somewhat like a human arm. Practical issues often support the "tool" development approach to arm prosthetics. For example, when a person holds an object with the elbow bent at 90 degrees, muscular action and expenditure of energy are required from the arm. Since the energy stored in a battery of an artificial arm would be depleted rather quickly under this approach, optional, nonphysiologic control choices like mechanically locking the elbow have to be made. In this document we have assumed that the prostheses to be controlled are basically "assistive tools." In other words, prostheses are the machine parts of a human-machine system and not a part of the human system- even though we want as much integration as possible.
It is advantageous for a prosthesis to move freely so that it can easily be put into the desired positions for operation. It may also be advantageous to control the rate of movement to the desired positions (the velocity). Once in position, it is often desirable to be able to control prehension force. Likewise, when a desirable position is reached, it may be advantageous to lock specific joints. Therefore, the variables to be controlled in arm prostheses of the kind under discussion are as follows:
- Prehension force
- The joint state (locked/unlocked)
There are many situations (e.g., pushing) where it is advantageous for a prosthetic arm to be completely rigid (all joints positively locked). There are other instances where the joints should be free (e.g., during walking). When we think of control we usually think of grasping or of positioning and lifting. However, the ability to make joints rigid or free is also an important function to be controlled in practical arm prostheses. In the future it may be advantageous to continuously control the impedance of joints from the free to the locked condition. However, that is not done in practical prostheses used by amputees today, and it will not be discussed here. It is currently practical to control the "free" and "locked" conditions, and this kind of joint impedance control will be emphasized in this chapter. It should be pointed out that friction joints, particularly for high-level amputees, do not function well because when an amputee wants to position a joint, the friction needs to be low, and when he wants the joint to remain fixed in position under load, the friction needs to be high. It is difficult if not impossible to meet both of these needs with a single friction setting. Therefore, locking/unlocking joints are often recommended, even though they may complicate control since the locking state must be controllable.
SOURCES OF BODY INPUTS TO PROSTHESIS CONTROLLERS
The human body can generate a variety of control signals that potentially could be used to operate prostheses. Childress has enumerated many of them. Practical inputs typically come from muscular activity (1) directly, (2) indirectly through joints, and (3) indirectly from by-products of muscular contraction (myo-electricity, myoacoustics, muscle bulge, and changes in muscle mechanical/electrical impedance). Although signals can be obtained from brain waves (electroencephalography [EEG]), voice, feet, eyes, tongue, and other places, these sources of control have not been shown to be practical for artificial limb control. A partial list of control options is included here, with concentration on options that are currently in common or partial use and on those that appear to have some future potential for use with practical limb prostheses. The options have been classified as two types: biomechanical and bioelectric/acoustic.
Biomechanical inputs of the kind described above have been used fairly extensively for the control of non-powered prostheses. These same inputs can be used with some powered prostheses. In fact, increased flexibility can be obtained for these inputs with powered prostheses since force/excursion requirements can often be considerably relaxed when powered components are used. The ways in which biomechanical inputs can be used for control are, for the most part, intuitive and will not be discussed here in detail. Basically, the force or movement of a body part (e.g., the chin/head) is used to move a mechanical switch lever, to activate an electronic switch, to activate a cable attached to a switch or instrumented element, to push on a pressure-sensitive transducer, or to otherwise operate some kind of position, force, or touch/proximity transducer.
There are many kinds of transducers that can detect biomechanical signals (force or excursion) and turn them into electrical signals that can be used for control purposes. It is not the intent here to discuss the many transducers that are available commercially. In fact, only those transducers that are used in presently available prosthesis control systems will be discussed. These are mechanical switches that require both force and excursion to turn on or off, pressure-sensitive transducers that change their resistance with force applied but with essentially no excursion (isometric), and excursion transducers that measure distance but with essentially no force required. Most of the suppliers of control systems, as described in Chapter 6C, supply switch controllers and myocontrollers. Universal Artificial Limb Co. supplies pressure transducers, and Hugh Steeper, Ltd., supplies an excursion transducer. Switches are applicable to most systems, and with a number of the control systems they can be used interchangeably. Care needs to be exercised when attempting to use transducers interchangeably (sometimes even switches) with control systems for which they were not designed or for which they are not specified as being compatible. Correct voltage amplitude, voltage polarity, electrical impedance, and electrical connections must often be observed when interconnecting transducers with control systems.
Bocker and push-button switches are commonly used switch types that can easily be operated by pressing against them with a body movement. Switches are easy to use, simple, and inexpensive. Also, their assembly into a whole prosthesis is fairly intuitive. Unfortunately, switch control is not always sufficient for good prosthesis control.
Switches also can provide more than one function from one source. For example, a frequently used pushbutton switch produces one function when pushed in a short distance and another function when pushed in a greater distance. In this way, the two functions of a powered prosthetic joint or prehensor can be controlled with the switch and activated by only one control source. Switch inputs can be arranged (with some electronics) so that multiple activations could be used to produce certain prosthetic functions. For example, a simple code (like a few of the simple letters of the Morse code) could be input to produce a specified output. This is not done. It is mentioned here only to hint at the wide variety of control schemes that are possible with simple switches and electronics. Many kinds of control systems and transducers could be used with prosthetics systems. With each system, questions must be asked. Are they reliable and simple to incorporate into a system, and more importantly, do they offer some or many of the desirable attributes of prosthetics control that have already been discussed?
The Integrated Nature of Prosthetics Systems
Limb prosthetics systems suffer from the kind of "reductionist" approach being pursued in this chapter. While it may be useful in one sense to break down upper-limb prosthetics systems into powered components, control systems, transducers, etc., and to only talk about control in this chapter, in another sense this may not be a good way to think about how to design a well-functioning system. It is the author's opinion that the best operating system for a given task needs to be built as an integrated whole and not through a "modular" approach where different componentry is cobbled together to create a total system. The modular approach has the advantage of providing great flexibility and novelty of system design. However, this approach will probably never be able to attain the highest functional goals that may be possible. Only systems that are designed from a more integrated standpoint may be able to accomplish this.
By definition, myoelectric control is the control of a prosthesis or other system through the use of "muscle electricity." In this kind of control, the control source is a small electric potential from an active muscle. This electrical potential is electronically processed and can be used to activate a switch controller or a proportional controller of power to an electric motor, which in turn drives the prosthetic system (e.g., hand or elbow). Muscle electricity is a by-product of muscular action, just as mechanical noise is a by-product of an internal combustion engine. The electrical signal may be picked up with electrodes on the surface of the body as well as by internally dwelling wire/needle electrodes or telemetry implants. Surface electrodes are currently the only practical way to pick up myoelectric signals for prosthesis control because in prosthetics applications the electrodes will be used daily for long periods of time each day. Hence, they must be benign to the skin and tissues. The surface method of detection of muscle activity is nicely illustrated in the standard electrocardiogram (ECG), which is the electromyogram (EMG) of the heart muscle. A gel-type electrolyte is usually applied to the skin during ECG procedures to lower the electrical resistance of the skin. However, with prosthesis control, gel electrolyte is not recommended on the electrodes because of possible skin irritation with long-term usage. Consequently, inert metal (e.g., stainless steel) electrodes are usually used in myoelectric prostheses. They are often called "dry" electrodes because of the absence of electrode paste (conductive gel). Actually, they are not "dry" in the normal sense of the word because the body's own perspiration serves as a reasonably good electrolyte for the electrodes and makes conductive pastes unnecessary.
Just as with an ECG, special care must be taken to negate the influence of interfering electrical signals from the environment (e.g., broadcast waves, fluorescent lights, motor arcing, power lines, etc.) that may cause the prosthesis to operate inadvertently. These potential interference signals may be many times larger than the myoelectric signal itself. A typical surface EMG may have a peak-to-peak amplitude of around 100 ÁV (0.0001 V), whereas the noise signals may be a thousand times greater in magnitude. The electrical noise can be eliminated, for the most part, by good electronic circuitry that features differential amplification, filtering, and thresholding and by good electrode positioning and design techniques. To reduce electrical noise pickup, the electronic amplifiers are often packaged together with the metal electrodes to make the connecting wires extremely short between the electrodes and the amplifiers. The reader should refer to Fig 6D-2. to see the electrodes, as drawn diagram-matically. When the electronic amplifier or the amplifier and processor electronics, as shown, are put into a single package with the metal electrodes on the outside, the whole package is often called an electrode; however, from a technical viewpoint it should be remembered that only the metal parts that interface with the user's skin are the actual electrodes. Amplifiers or other circuitry at the electrode site are part of the electronic amplifying and processing system.
It is impossible to cover myoelectric control comprehensively in this chapter. The characteristics of myoelectric signals and the processing of myoelectric signals for use in prosthesis control have been described extensively in many places. Good technical sources for information in this area may be found in a review of myoelectric control by Parker and Scott and in Bas-majian and DeLuca's discussion of myoelectric signals. Scott has written an elementary introduction to myoelectric prostheses, including control, and Scott and Childress have prepared a comprehensive bibliography concerning myoelectric control of limb prostheses.
The use of myoelectric control in arm prostheses has greatly increased in the United States and elsewhere during the last decade. Consequently, some may consider this technique a result of "space age electronics." In reality, the first myoelectric control system was built in Germany about 1944. The physical concept is therefore nearly 50 years old, older than the solid-state electronics that made the method ultimately practical. The early German system and an early British systemwere designed with vacuum tube electronic technology. British scientists were instrumental in advancing the concepts of myoelectric control early on and constructed some novel circuitry. Soviet scientistswere the first to design a transistorized myoelectric system that could be carried on the body. Collaboration between a German company, Otto Bock, and an Austrian hearing aid company, Viennatone, led to the first transradial myoelectric system that could be commercially purchased in the United States. Many other commercial myoelectric systems have followed (see the current listing in Chapter 6C).
Although myoelectric control will not be discussed here in great detail, it seems appropriate to discuss this commonly used control method in a general way so as to give the reader a sense of what it is about. Since there are no systems within our common daily experience that are analogous to myoelectric control, it seems appropriate to describe it more fully than was the case with biomechanical control approaches, which are more intuitive.
Electricity from skeletal muscles can be created by voluntary muscle action. In fact, this voluntary control is one of the excellent attributes of myoelectric control. A myoelectrically controlled system will only work when the amputee wills it by voluntary muscle action. Such a system is immune to influence from external forces, prosthesis location, or body position/motion. Similarly, except for very exceptional cases, the prosthesis should be free from influence by environmental electrical noise.
The myoelectric signal itself is a rather random-shaped signal that comes from the spatial and temporal summation of the asynchronous firing of single motor units within the muscle. It is a kind of electrical interference pattern resulting from the electrical depolarization of thousands of muscle fibers (perhaps several hundred per motor unit for typical forearm muscle action) when they are activated by neurons. This kind of random-like electric wave can only be described statistically because its amplitude and frequency are constantly varying, even when a person is holding his muscular action as constant as possible. However, one can use a "rule of thumb" to remember the general range of amplitude and the dominant frequency of a typical surface signal. The rule of thumb is to remember the number 100 for amplitude and for frequency- 100 |xV for amplitude and 100 Hz for frequency. A typical surface EMG amplitude on the forearm, under moderate muscle action-which can be measured in a number of ways (peak to peak, root mean square (RMS), etc.)-is often in the neighborhood of 100 ÁV, or on the order of a million times less than the voltage of electrical wiring in American homes. Of course, this voltage can usually be made larger by increased muscle action, or it can be reduced all the way to zero when the muscle is inactive. The frequency components of the EMG that have the most energy are in the neighborhood of 100 Hz (cycles/sec). There is very little energy in a surface EMG above about 400 Hz.
It is frequently desirable in electronic design to amplify the voltage of the surface EMG up to a level of from 1 to 10 V. Consequently, we can see by our "rule of thumb" that an amplification of 10,000 to 100,000 is needed (1.0/0.0001, or 10.0/0.0001) to accomplish this increase. To avoid noise amplification as much as possible, band-pass differential amplifiers are used so that voltages common to the two inputs (common-mode voltages) are rejected and so that amplification is most effective for frequencies around 100 Hz. No amplification is necessary above about 400 Hz for control purposes since the signal above this frequency is relatively low. It should be noted that additional bandwidth is necessary for instrumentation purposes (e.g., up to 1000 Hz). Frequencies below about 10 Hz are frequently not amplified to any extent so as not to amplify slow polarization voltage changes that may occur over time at the electrode-skin interface, which may be of special importance with "dry" electrodes.
It should also be noted that a myoelectrically controlled prosthesis can only function in its normal way when all the electrodes are positioned properly on the body. All electrodes should remain in contact with the skin at all times during prosthesis usage. If electrodes lose contact with the skin, a lack of control or interference may result. For this reason it is important for the prosthetist to fabricate a diagnostic prosthesis with a clear plastic socket that permits the electrodes to be observed while the prosthesis is used in various positions and under various prosthesis loading conditions. The socket needs to be designed so that the electrodes maintain contact with the skin for all reasonable external load applications and for all reasonable prosthesis positions and movement velocities.
The body acts as an antenna and picks up electrical noise from the environment. Consequently, touching the exposed electrodes with the fingers-so called "tipping"-introduces electrical noise through the fingers to the electrodes and into the electronics. There are no myoelectric signals in the fingertips. Also, this response should not be interpreted to mean that the electrode is a touch sensor or a pressure sensor during regular use; it is not. It merely means that when touched the myoelectric system responds to the stray electrical noise present on the fingertip. "Tipping" the electrodes is often used as a way of demonstrating the general action of the prosthesis when it is not on the body. However, it must be remembered that an expected response to touching the electrodes does not necessarily mean that the myoelectric system is completely functional. Malfunctioning amplifiers may still respond to "tipping" even when they no longer function correctly as myoelectric amplifiers. Therefore, a correct "tipping" response is a necessary but not a sufficient test to determine whether a myoelectric prosthesis is functioning properly.
In a myoelectric system, amplification is followed by electronic processing that usually turns the myoelectric signal, an ac potential, into a dc potential of a given polarity (positive in Fig 6D-2.). The envelope of this dc potential goes up and down as the myoelectric signal increases or decreases in amplitude-as the muscular action increases or decreases. Electronic logic circuitry can be designed such that if the dc potential is greater than some threshold voltage (e.g., 1.0 V), then the circuit will turn on an electronic switch that allows electric power to flow to the prosthesis motor. Therefore, the result of contracting a muscle to a certain level results in power delivery to the driving motor of the hand or arm. If the dc potential falls below the threshold, the power to the motor is turned off
It should be noted that in myoelectric control it is the voltage and current from the battery that provide power to the motor, not the electricity from muscles. The myoelectric signal is used only for activation or control purposes. The system illustrated in Fig 6D-2. represents the essence of myoelectric control of a prosthesis motor-a kind of generic myoelectric control module. In actuality, the electronics of myoelectric control systems from each manufacturer take on different forms and designs. Some have circuits that enable the power to be applied to the motor in a manner proportional to the myoelectric signal amplitude. Some can turn the motor on and also reverse its direction of action (polarity/rotation) while using only one myoelectric control site. Others use two or more myoelectric control sites to effect action of a motor or motors. Fig 6D-3. shows a typical transradial myoelectric prosthesis and a generic design for a two-site, two-function myoelectric control system for it.
Myoelectric control of a hand or other prehensor is particularly applicable to transradial amputation levels since people with acquired amputations usually have a "phantom sensation" of their missing hand. When they think of moving their phantom hand, the muscles remaining in their limb are naturally activated. Therefore, it is possible to relate original finger extensor muscles with "opening" of the prosthetic hand (often in conjunction with use of wrist extensor muscles) by placing electrodes on the skin near these muscles. Likewise, the original finger flexor muscles can be used (usually in conjunction with wrist flexor muscles) for the signal site to "close" the prosthetic hand. As a consequence, there can be a rather natural relationship between thinking about operating the phantom limb and actual operation of the hand prosthesis. Also, normal elbow control by the transradial amputee allows him to move the hand in space and to have proprioception concerning where it is with respect to the body, how fast it is moving through space, and what external forces are acting upon it. Consequently, transradial myoelectric prosthesis control is usually performed well.
Myoacoustic signals (auditory sounds when muscles are active), a phenomenon observed long ago but only recently reinvestigated in much depth, have been shown to have potential for the control of prostheses.Myoacoustic control systems are very similar in structure to myoelectric systems, and there does not appear, at this time, to be any compelling reason to move from myoelectric control to myoacoustic control. Myoacoustic controls primary advantage over myoelectric control could be that the acoustic sensor does not have to be in direct contact with the skin. Its main disadvantage concerns potential difficulty with elimination of extraneous mechanical noises. When a prosthesis strikes an object in the environment or rubs against something in the environment, large mechanical vibrations can be created. The elimination of this unwanted acoustic noise may be more difficult with myoacoustic control than it is with unwanted electrical noise reduction in myoelectric control systems.
Neuroelectric control, where microelectrodes interface directly with nerves and possibly with neurons, remains a control possibility that may have future applications. This method of control requires indwelling components of some kind (e.g., telemetry implants) because neuroelectric signals are, in general, too weak to be picked up on the surface of the skin. The method has the potential advantage of multiple-channel control and multiple-channel sensing because there are many motor and sensory neurons associated with each nerve. Nevertheless, the method is experimental and only has "potential" for practical applications. Nervous tissue is rather sensitive to mechanical stresses, and so it may be difficult to maintain long-term neuroelectrodes. The practicality and effectiveness of this kind of human-machine interconnection will remain an open question until it can be tried extensively.
Another surgical possibility with nerves is to surgically connect the cut ends of nerves to prepared muscle sites. This has been suggested by Hoffer and Loeband experimentally investigated in basic studies of animal preparations by Kuiken. The concept, for example, would be to take a muscle like the latissimus dorsi, which may not be a functionally critical muscle for a shoulder disarticulation amputee, remove its normal innervation, and reinnervate it at multiple places with nerves that formerly went to the hand and forearm. The muscle, after reinnervation, might be a good source of multiple myoelectric sites or other kinds of control sources for prosthesis control. This technique has the benefit of not requiring implants. Andrew (see Chapter 9B) has apparently fitted some transhumeral amputees with myoelectric control who had had successful nerve transfer following brachial plexus injury.
The decade from 1965 to 1975 was one of unprecedented research on the control of artificial limbs. The research, particularly that conducted in Europe, was stimulated by limb absences at birth that resulted from use of the drug thalidomide during pregnancy. That period of activity and research ferment was also marked by the excitement that resulted from the practical introduction of myoelectric control during the mid-1960s. The Swedish Board for Technical Development sponsored a workshop on control of prostheses and orthoses in 1971, and the proceedings of that meeting is a landmark publication on prosthesis control research.
THE ROLE OF SURGERY IN THE CREATION OF CONTROL SITES/SOURCES
We know that amputation surgery is very important to the clinical outcome of prosthetics fittings. Nevertheless, this area of prosthetics is perhaps not emphasized as much as it should be. Surgeons can play an important role in assisting with control of limb prostheses. Unfortunately, the number of surgeons with an active interest in amputation and amputation issues seems relatively diminished today as compared with the years immediately following World War II. For control of limb prostheses to advance along a broad front, advancements in surgery and surgical techniques are as necessary as technical advancements. In fact, technological advancements and surgical advancements in prosthetics should be integrated, synergistic activities. Techniques in orthopedic surgery, vascular surgery, plastic surgery, and neurosurgery have advanced rapidly over the last 20 years; unfortunately, many of the new techniques have not had as much impact on prosthesis control as they could have had if surgeons, pros-thetists, and engineers had consistently worked together on limb control problems. Surgical procedures in general are handled in other parts of the Atlas, but objectives of surgery in assisting with control sites and function are considered here.
Bones and Joints
As a general rule, the surgeon should try to save joints and bone length, consistent with good medical practice. This procedure, in general, leads to improved prosthesis control. With transradial amputations it is usually desirable to make the limb as long as possible. Wrist disarticulations are desirable since they conserve natural supination-pronation of the forearm, provide contours for prosthesis suspension, and create a force-tolerant distal end for the limb. The elbow joint should always be saved, if possible, since it greatly enhances prosthesis control, just as saving the knee enhances lower-limb prosthesis control. Bone lengthening might be considered for increasing the length of very short limbs, where it is practical. Decisions concerning saving the wrist joint, if all fingers have been amputated, have to be made on an individual basis. In the future, if advanced finger components are available for the partial-hand amputee, it may be useful to save the wrist because of the highly desirable movements it provides for positioning an artificial prehension component. At present, the range of fitting options that can be achieved are limited at this amputation level (e.g., passive/cosmetic prosthesis, opposition post, HandiHook, Robin-Aids Hand) because of size and length constraints. Some partial-hand amputees who have no fingers or thumb decide that they want to have their limbs revised (shortened) to the wrist disarticulation level so that they can easily be fitted with standard electric hands and myoelectric control.
From a control viewpoint, transhumeral amputations seem to follow guidelines similar to those for the transradial amputation. Elbow disarticulations conserve humeral rotation, can be used to aid prosthesis suspension, and provide a force-tolerant distal end. Long transhumeral limbs often obtain good control of prosthetic elbow flexion by using glenohumeral flexion; however, if a disarticulation is not possible, the length should usually be reduced enough to accommodate elbow mechanisms without compromising function. Mar-quardt has used angle osteotomies of the distal end of the humerus to improve mechanical coupling between the humerus and the prosthesis so that the humeral rotation of the prosthesis is readily controlled by natural humeral rotation.
A control viewpoint suggests that the surgeon should attempt to save a short humerus if it will be voluntarily mobile because a mobile short humeral neck can be used to activate control switches or to push against pressure-sensitive pads. Muscles attached to it may also be used for myoelectric control purposes. If amputation above the elbow is performed after brachial plexus injury, it is often helpful to have the flail humerus fused with the scapula at the glenohumeral joint. In this way the humeral section can be controlled, to some extent, by action of the scapula.
Soft-Tissue Conservation and Reconstruction
Surgeons should conserve residual muscles that might be used for myoelectric sources if the conservation is consistent with good medical practice. Myoplasty procedures that connect antagonist-agonist muscle groups at the distal end of the amputation site are often used in order to keep the muscles in a dynamic, somewhat natural working relationship. It is felt that this promotes the possibility of good two-site myoelectric control from these muscles. Myodesis is sometimes performed. Single muscles that may have no functional purpose after amputation but that can be voluntarily activated should be attached to a reaction point and saved for possible use as a myoelectric control site. For example, if part of the deltoid muscle can be conserved after arm disarticulation, its free end may be attached to the torso for possible use as a myoelectric control source for an arm prosthesis.
Although tunnel cineplasties have not been used much in the United States since the 1950s, they offer a unique way for surgeons to create control sources. New surgical techniques and the wide availability of powered prostheses may lead to a revival of this procedure. The technique is being reconsidered in Europe. In recent years Baumgartner, Biederman, Krieghoff et al., among others, have written about the utility of this control technique. In the United States, Leal and Malone successfully fitted a transradial amputee who already had a standard biceps tunnel cineplasty with an electric hand that was switch-controlled from the cineplasty site. Liicke et al. have discussed the use of cineplasty in connection with modern electronic prosthesis technology. Beasley introduced a new cineplasty-like, "tendon exteriorization" procedure that shows much promise. Tendon exteriorization does not traumatize the muscle itself and therefore is thought to have minimal influence on a muscle's circulation and neurologic mechanisms. This procedure demonstrates the possibility for surgical creation of a number of such control sources on the forearm that could, in the future, enable amputees with long transradial limbs to gain coordinated control of individually powered prosthetic fingers. Also, the surgical creation of a number of new tunnel cineplasty control sources on the torso may be particularly desirable for the high-level bilateral amputee who needs multifunctional control but who has limited control sites without such surgical intervention. Direct muscle control through tunnel cineplasties is particularly attractive in both cases because of the proprioception they naturally provide to the user. This is thought to be a particularly desirable feature for obtaining good control of multiple prosthetic functions without too much mental effort given over to the control process by the user. Powered prostheses make it possible for tunnel cineplasty control sources to be used even when they can develop only small forces or small excursions. It appears that the combination of powered prostheses and electronic position control systems, in conjunction with new surgical techniques and procedures, may open up a new era of control based on the older, but still vital ideas of tunnel cineplasty.
Adherence of the skin to underlying muscle is a less direct method of using a muscle as a control source. Skin adherence brings about skin motion when muscular contraction causes movement. This method of control has been demonstrated by Seamone et al.
Surgical transfer of muscles to the amputated limb is possible for improvement of arm control, but this has not been done in large enough numbers for generalizations to be made about the utility and indications for muscle transfer procedures. Joint control, myoelectric control, or tunnel cineplasty control may all be possible applications for muscle transfers. Finally, it should be mentioned that the Krukenberg procedure remains a viable method to allow direct prehension control and can also be used for control of powered transradial prostheses. It presents options. Some users may choose to use the Krukenberg limbs in the privacy of their homes because of the good sensory and motor qualities, but they may prefer to use prostheses over their arms when they are in public venues. Even though this procedure has normally been used primarily with blinded bilateral hand amputees, the procedure may, in certain circumstances, have applications with sighted and unilateral amputees. Activation of pressure-sensitive transducers by the Krukenberg limb is one way to use it to control a hand. Myoelectric control is another option. Finally, it may be possible to use the concepts of extended physiologic proprioception with the Krukenberg limb in order to gain improved control of a powered prosthesis.
It is important to remember that surgical procedures may need to go beyond just the original amputation in an attempt to create a limb that will be functional and easily fitted and that will not cause subsequent problems for the amputee. Surgery can be very beneficial in assisting with the control of prostheses. This is of particular importance for high-level bilateral amputations, but may also be important for less difficult cases where a number of control sources are needed for multifunctional prostheses. Sometimes the surgical procedures designed to assist with prosthesis control can be performed at the time of original amputation, but often such surgical intervention will necessarily need to be done at a later date.
CONTROL OF PROSTHESES FOR SINGLE (UNILATERAL) LIMB LOSS
While fittings of unilateral amputees are technically much more simple than bilateral fittings, evaluation of the results of unilateral fittings may be more difficult than for bilateral cases. This is because the unilateral amputee has a physiologic arm and hand that, if normal, can accomplish almost any task. Therefore, the prosthesis at best serves only in an assistive mode. Some amputees may want it primarily for appearance purposes. Others may incorporate it extensively into their activities and body image, while some may have only specific functions (jobs or sports activities) for it. Many unilateral amputees decide not to wear a prosthesis. Professionals connected with prosthetics fitting need to support amputees' decisions about the use of prostheses. Nonetheless, they also need to be able to inform the amputee concerning what kind of fitting can potentially serve him in the best way by taking into account the many factors involved. The prosthetics fitting of upper-limb amputees is partially an iterative process because amputees cannot know what problems they will face until they actually use a prosthesis and experience it in their natural environment. Also, an amputee's true feelings and desires may take time to mature and to emerge. Likewise, an amputation often leads to job changes and other changes that require time to be sorted out. The prosthesis takes its place amid many life changes, and this makes initial prescription difficult. These factors suggests that diagnostic and temporary prostheses may be very useful for initial and early fittings.
Control of Unilateral Transradial Prostheses
Transradial prostheses may be controlled successfully in many ways. Cable-controlled voluntary-opening and voluntary-closing prehensors (nonanthropomorphic) both work well with transradial amputees, although cable-controlled mechanical hands are generally inefficient. When myoelectric control first became available, the conventional wisdom was that it might have more important usage with higher-level amputations and was not so important for transradial prostheses. However, experience has shown that myoelectric control works admirably at this level. In fact, it has its greatest application with the transradial amputee. A myoelectrically controlled transradial prosthesis is shown in Fig 6D-3.. If possible, it is preferable to use two myoelectric sites to control the two functions (closing/opening) of the hand because this gives the operator direct control of each function. This kind of control can become rather subconscious in nature for some amputees. Of course, the prehensor can be various electric hands or various nonanthropomorphic electric prehensors as described in Chapter 6C.
Single-site, two-function control is quite acceptable for amputees who do not have two good myoelectric sites. It has been used effectively with youngsters (e.g., the New Brunswick system) and with adults. In like manner, the single-site, single-function myocontroller of hand opening with automatic powered closing (the St. Anthony control circuit, the so-called "cookie crusher'' system) has been shown to be effective with very young children who are born with limb absences. This is similar to the single-site, single-function myocontroller for voluntary-opening prehensors (Hosmer's Prehension Actuator), which can be used to provide powered operation to a variety of voluntary-opening devices that traditionally have been controlled through cables and body power.
A pair of single-site, two-function controllers can be used to control four functions of powered hand opening/closing and powered pronation/supination for the short transradial limb. In general, supination/pronation is not necessary for most unilateral amputees unless a particular vocation or avocation demands it. Powered supination/pronation adds weight distally and also adds complexity. Sockets and apparatus that allow natural supination/pronation (body powered) from residual movements of the amputated limb are recommended, when possible.
Proportional control has been shown to be effective by Sears and Shaperman. This is intuitively understood; however, if powered prostheses have slow dynamic responses (e.g., hands close or open at slow rates), then proportional control is not necessary for effective control; on-off control is sufficient. Rapidly moving prostheses that have maximum angular velocities greater than 2.0 to 3.0 radians/sec (-115 to 172 degrees/sec) normally will require proportional control, although few devices with this speed are currently available. At some operating velocity, accurate control of position becomes impossible for the human-operator using on-off velocity control (as noted previously in the discussion on control of electric-powered automobile windows).
Electric switches in series with Bowden cables can be used to control powered hand prostheses. However, this kind of control probably should be avoided if possible because a prosthesis controlled in this way by a transradial amputee has the disadvantages of cable control (harness and limited work envelope) along with the disadvantages of powered hands (weight, battery, mechanical complexity). A possible exception would be when the switch is operated from a tunnel cineplasty; however, it seems likely that position control, as with the Steeper hand controller (see Fig 6C-2) or with some other kind of position control mechanism, would be preferred for hand control through a tunnel cineplasty. In this situation, the muscle position would determine hand opening position. If muscle velocity could also be related to prosthesis velocity and muscle force to prosthesis gripping force, then the muscle might provide some proprioceptive sense to the user.
Evaluation of the Effectiveness of Control Approaches
As the reader can see from the transradial unilateral situation, there are many ways to fit amputees from a control viewpoint and from a component viewpoint. Which way is best? Is there a "best?" With higher-level amputations the question becomes even more important because the fitting options are multiplied. Powered and nonpowered components at multiple joints coupled with various control schemes yield many possible fitting combinations. How can the various options be evaluated? Of course, everyone has his own opinions about what is best, which may be subject to change or may not. (Note that although this question can be raised in all areas of prosthetics, it just becomes very obvious in high-level fittings.) Unfortunately, the question can only be answered quantitatively for particular criteria that are arbitrarily selected and through large-sample studies. It can be answered subjectively by the prosthesis users, but again a large-sample study would be required. Such studies are difficult to fund and conduct, and the results may be equivocal. At present, a combination of "rules of thumb" based on experience and clinical judgment usually determines the initial prosthetic approach.
Surgeons have the same problem in evaluating many surgical procedures. Their use of case studies and retrospective analysis of results suggests the need for studies of this kind to be performed by professionals in prosthetics who have reasonable caseloads of upper-limb amputees. It is a partial solution, at best, to the evaluation of effectiveness of control techniques and methods. The author proposes that an alternative evaluation approach might be to measure control methods against the "Desirable Attributes of Prosthesis Control" presented in the first part of this chapter. If they have many of the attributes, they would rank higher than if they do not. This may be difficult to quantify, and not everyone will agree upon the desirable attributes, but it may be a viable first approach to the problem. Another option is to base evaluation of control approaches on the basis of a theoretical construct. Fittings with close correspondence to the theory would get higher ratings than other fittings. These ratings would be incorrect if the theory was incorrect, but if the theory was correct, they would be valid. A theoretical construct proposed by the author is discussed later in this chapter, and simple examples of evaluations based on the construct are presented.
Control of Unilateral Transhumeral Prostheses
It seems appropriate to present the most common control approaches currently used in prosthetics practice for powered transhumeral prostheses along with an emerging approach of interest to the author. As with transradial amputees, transhumeral amputees with relatively long limbs can function well with totally cable-operated, body-powered prostheses. Several other approaches are common:
- For transhumeral amputees with long residual limbs, the hybrid approach of a cable-operated, body-powered elbow along with myoelectric control from the biceps (closing) and triceps (opening) of a powered pre-hensor (hand or nonhand) is a very functional fitting approach. This approach has been used effectively in Europe for almost 25 years. Billock has used this technique effectively with many people. It is a relatively simple approach-technically comparable to a transradial myoelectric fitting. This kind of fitting is shown in Fig 6D-4.,A. The hybrid control/power approach has reasonable proprioceptive qualities and allows simultaneous coordinated control of elbow and prehensor function. It avoids the problem of prehensor opening during forearm lifting against a load, which is a problem with a cable-operated elbow if the cable is also used to operate a voluntary-opening (spring return) prehensor. The author feels that myoelectric control of prehension, in this case from the biceps and triceps, is somewhat natural because gripping objects strongly often involves the contraction of muscles quite distant from the hand. The relationship between prehension and muscular contraction has been called the "myopre-hension" concept.
- Hugh Steeper, Ltd., has a body-powered elbow that is designed for a hybrid control approach to trans-humeral fittings. The mechanical elbow has an electrical switch in it that is connected with the elbow locking mechanism. When the elbow is unlocked, the electrical switch is open, and when locked, the switch is closed. This allows a single cable to operate a servo-controlled hand and also the elbow, without interaction. When the cable is pulled to operate the unlocked elbow, the electrical connection to the hand is turned off. When the elbow is locked, the connection to the hand is on, and pulling on the cable operates the hand through the position servo control system. Another way to use this elbow design is to place a two-position switch in series with the cable that controls the elbow. When the elbow is unlocked, cable operation is normal. When the elbow is locked, pulling the cable lightly will activate the first position of the switch and close the hand. Pulling the cable with greater force will activate the second position on the switch and open the hand. In both cases the idea is to reduce the number of control sources needed. However, simultaneous control of both functions is impossible with this control approach.
- An alternate but similar approach is to use a powered elbow in place of the body-powered elbow but to control it in a similar way: using the cable to operate a position servomechanism controlling the elbow. This approach, shown in Fig 6D-4.,B, is a kind of "boosted" cable control. Since the cable is directly connected to the elbow's output position, the body's position cannot get ahead of the corresponding position of the elbow and forearm. Therefore, it is a form of "unbeatable" position controller that is similar in operation to automobile powered steering, mentioned earlier in this chapter. The approach is based on D.C. Simpson's principles of extended physiologic proprioception.Heckathorne et al. have reported on this technique
for a clinical fitting. The advantages are that proprioception is maintained even while using a powered elbow and that the force and excursion necessary to operate the elbow can be matched to the amputees force and excursion capabilities. The principles and details behind this particular control approach have been described by Doubler and Childress.
- For transhumeral amputees who cannot operate a body-powered elbow well (e.g., have trouble with the locking and unlocking function), a powered elbow can be used, often myoelectrically controlled (biceps-flexion, triceps-extension). The prehensor can be cable controlled and body powered. This is thought to be an effective work prosthesis if a totally cable-driven system cannot be used. It is an approach that has been promoted for use with the Liberty Mutual electric elbow.
- For transhumeral amputees who do not want to use the harness needed for cable control or who cannot tolerate a harness (e.g., because of skin grafts) or for amputees with a relatively short limb (weak glenohu-meral leverage), the controls can be completely myoelectric, as with the Utah arm fitting shown in Fig 6D-5.. This is a two-site myoelectric control system that can be used to control the elbow proportionally. If the elbow is held stationary at a position for a short period of time, the elbow automatically locks, and this action transfers the myoelectric proportional control to the hand. A quick cocontraction of the biceps and triceps muscles is used to transfer control back to the elbow. This is a form of two-site, four-function control in which all functions are not directly accessible. Control can be alternated between the hand and the elbow.
Control of Unilateral Shoulder Disarticulation Prostheses
Unilateral shoulder disarticulation amputees often choose not to wear a prosthesis. Some prefer to wear a lightweight passive prosthesis that is free to swing comfortably during walking and that can be easily positioned (passively) for placing its cosmetic hand in their lap when they sit. Light, passive holding by the cosmetic hand may provide some utility. Body-powered prostheses are marginally effective at this level of amputation-when the contralateral limb is fully functional. The user often has somewhat limited force and excursion when compared with amputees with mid to long transhumeral limbs, and a body-powered system may be difficult to operate. A powered prosthesis (e.g., with electric elbow and electric hand or other powered prehensor) may also not be desirable at this level of amputation because, again, the functional gains provided are likely to be marginal when the opposite limb is fully capable. Powered limbs also add undesirable weight, a detrimental factor in this kind of fitting.
If a powered limb should be fitted for this amputation level, it would likely be used mainly as an assist to the normal contralateral limb, primarily with its prehensor acting as a conveniently located viselike holding mechanism (e.g., holding a bottle while the other hand takes off the cap). An electric elbow and electric prehensor could be used in conjunction with friction or manually locking wrist rotation, friction or manually locking humeral rotation, and friction or a manually locking shoulder. The user would preset wrist rotation, humeral rotation, and shoulder position with his capable limb and would use the control system to position the elbow joint and operate the prehensor.
A convenient control scheme for this situation would be movement or force from the shoulder on the amputated side. If support for the arm can come from the torso, then the structure can be contoured so that the shoulder is free to move up and down and back and forth, to a limited degree, within the prosthesis. This relatively free motion can be used effectively for control. The author feels that position servo control of the elbow, interfaced with up and down movement of the shoulder (up for elbow flexion), proportional force control of the prehensor mediated through shoulder protraction (the prehensor closes with a force proportional to the shoulder force), and retraction movements against pressure-sensitive transducers would be a desirable control scheme. However, many other schemes would work effectively, and the differences in overall performance of the unilateral amputee, as a whole, would probably not be discernible with many other control approaches (e.g., mechanical switches operated by the shoulder movements), particularly with most currently available electric elbows and prehensors.
Control of Bilateral Arm Prostheses
The fitting problems become dramatically different when both arms are missing. We will not address the fitting issues involved with all of the various combinations of bilateral limb loss. If we only consider transradial and transhumeral amputation combinations, we have 4 combinations but only 3 different varieties (2 combinations are equivalent to each other, left and right) of amputation conditions (bilateral transradial, bilateral transhumeral, and transhumeral-transradial). If we add "long" and "short" to the classification of each amputation level, we have 16 possible combinations, with 10 different varieties (12 combinations have like equivalents, left and right) of amputation conditions. Therefore, the number of variations can be large if several different amputation levels of each limb segment are considered and much larger if associated movement limitations or pathologies of each limb segment are also included. We will consider only a few conditions of the many varieties of amputation conditions possible and will concentrate on general principles for the fittings rather than on specific details.
Amputees with bilateral long transradial limbs can effectively control a wide range of prostheses from cable-controlled voluntary-opening hooks to bilateral myoelectric hands. Attempts should be made to maintain the physiologic pronation-supination remaining-on both sides. Passive (friction or locking) wrist flexion will be useful, at least on the dominant side. It might be useful to use two kinds of prehensors, one voluntary closing and one voluntary opening, although the author has not seen this done. This would provide body power on both sides, but the prehensors would be complementary in function. The voluntary-closing prehensor would enable high prehension forces to be developed, and the voluntary-opening prehensor would permit relaxed, unattended prehension. Another possibility is to use different kinds of prehensors and different control schemes on each side. Body control with passive wrist flexion could be fitted on the dominant side with a voluntary-opening or -closing prehensor. A transradial myoelectric hand prosthesis (or nonanthro-pomorphic prehensor, e.g., Greifer or Synergetic Prehensor) with socket provision to capture residual forearm rotation could be fitted on the nondominant side. This would give the wearer the advantages of both kinds of systems-the precision prehension capabilities of many hook prehensors along with good proprioception from the cable-operated control system and the power prehension of an electric prehensor along with the large work envelope that is possible with a myo-electrically controlled prosthesis. The two systems should complement each other, and the controls should be as independent as possible. There are many options, and the one chosen will be highly dependent upon the needs and preferences of the user. Powered hand prostheses may be used bilaterally with aesthetic advantage but often with functional disadvantage because the hands are usually limited to one prehension pattern (palmar prehension) and because their bulk makes it difficult to use them in constricted spaces (e.g., pockets).
If both arms have transradial amputations, one long and the other short, the long limb would normally be fitted as the dominant limb. Again, as before, a wide range of fitting possibilities are possible. An all cable-controlled system with hooks can be very effective, as demonstrated by so many amputees who generally develop exceptional arm/prehensor skills. A variant of the complementary body-powered, externally powered system discussed in the previous paragraph may also be useful with this set of amputation levels. Powered supination-pronation on the short, powered, nondominant side should be considered. Similar control procedures are usable with the bilateral, short, transradial amputation condition. However, passive rotation of the prehensor should be added (along with the wrist flexion) on the body-powered, dominant side.
A person with a combination of transhumeral and transradial amputations can also be fitted well with body-powered, cable-controlled systems. The functional dexterity possible at this level with this kind of control can be extraordinary. People fitted in this manner fly airplanes-just one way they manifest their excellent control capabilities. The transradial side would normally be considered the dominant side fitting. Another scheme, if the transhumeral stump is reasonably long, would be to use cable control on the transradial side and a cable-controlled, body-powered elbow on the transhumeral side in conjunction with myoelectric control of an electric prehensor (as described for the long unilateral transhumeral amputation). When the transhumeral limb is short in this situation, a powered elbow should be considered.
Two transhumeral amputations frequently lead to the use of external power on one side or the other, although totally body-powered, cable-controlled systems can be functional at this level. The group the author works with at Northwestern University and at the Rehabilitation Institute of Chicago believes that there is merit in fitting these amputees with a body-powered, cable-controlled system on one side, usually the side with the longest residual limb but possibly on the side of the individual's original dominance, if the residual limb length is adequate. A single cable control of four body-powered functions has been found to be very functional. This is a technique pioneered by Mr. George Robinson at Robin Aids Prosthetics (Vallejo, Calif) and applied there currently by Mr. James Cay-wood. Their system has been redesigned somewhat to make it more modular and easier to apply. The concept is to use the primary cable control to open the voluntary-opening prehensor (hook) and to flex or extend the elbow (when it is unlocked) as is the usual case. However, with four-function control, two additional functions that can be locked (like the elbow) are added. These are a locking/unlocking wrist rotator and a locking/unlocking wrist flexor. As long as the elbow, wrist rotation, and wrist flexion units are all locked, the primary cable will pull the voluntary-opening hook prehensor open. If the elbow is unlocked, this cable controls flexion/extension. If the wrist rotation unit is unlocked by pushing a lever mechanism (e.g., lever on the forearm), activating the primary control cable supi-nates the prehensor if all other joints are locked. If the primary cable is relaxed, a spring connected to the wrist rotator pronates the prehensor. Therefore, the rotator can be unlocked, positioned to a new rotation angle by the primary control cable, and then locked again. The wrist flexor operates in a similar way. The amputee pushes a lever to unlock it (e.g., chin-operated lever). A rubber band causes it to move toward its extended position. Pulling on the primary control cable flexes the wrist unit. It may be locked at the desired position (in this case, three positions). If both the wrist rotator and the wrist flexion unit are unlocked, they move together. The prehensor will move to the extended and pronated position if the primary cable is relaxed. Pulling this cable under this condition brings the prehensor to a flexed and supinated position. This technique is shown on the right prosthesis of Fig 6D-6. (note the lever on the medial aspect of the forearm and the chin lever that is obscured under the shirt). The technique is also shown in Fig 6D-7. on the right prosthesis, except that in this case one chin lever locks/ unlocks the elbow and the other lever locks/unlocks the wrist rotator. The flexion wrist is unlocked by pushing the lever on its medial side in this case.
The single control cable, four-function control approach allows the bilateral transhumeral amputee to independently position joints of the arm and to lock them into position-an operation that is very helpful for the bilateral amputee. A body-powered elbow and myo-electrically controlled prehensor can be fitted to the nondominant side if the residual limb is fairly long. An electric elbow with myoelectric or rocker-switch control may be useful if the limb is short. This kind of fitting is illustrated in Fig 6D-6..
Short transhumeral and bilateral shoulder disarticulation amputations are cared for in our center with similar components as in the previous case, but the control methods may vary if a short transhumeral limb can be used as an EMG control site or if it can be used as a control source to push against pressure-sensitive transducers. As in the previous case, we like to use a four-function, body-powered, cable-controlled system on the dominant side. The nondominant side is fitted with a powered elbow, a powered prehensor, and a powered wrist rotator. The wrist rotator and the powered prehensor are controlled by chin movement against rocker switches. The elbow is controlled by a two-position pull switch that is activated by shoulder elevation. This kind of fitting is illustrated in Fig 6D-7. . Heckathorne et al. have described the complementary function of bilateral hybrid prostheses of this nature. The user can don and doff the prosthesis independently and uses it effectively in activities of daily living. Nevertheless, he has found it useful to also modify his home environment to simplify function.
Using a totally body-powered system on one side with a totally electric system on the other allows the two systems to be effectively decoupled from a control standpoint. In other words, forces and motions to activate the body-powered side do not activate the electric system on the opposite side. Likewise, operation of the electric prosthesis does not activate the body-controlled system. This automatic decoupling allows the amputee to concentrate on the prosthesis he is operating without having to consider both simultaneously. Initially, all joints except at the shoulder had positive locks, so the user does not have to worry about them slipping under loads. The shoulders had friction joints that were pre-positioned and set to high friction (see Fig 6D-7. ,A). Later the shoulder joints were converted to positive locking joints (MICA, from M. Collier, Longview,Wash) that have positive locking/unlocking in flexion-extension and friction in abduction-adduction. One of these is shown installed on the right shoulder in Fig 6D-7. ,B. The three locking levers and two electric rocker switches shown in this figure are operated easily and unobtrusively by the amputee. Chin control appears to be integrated nicely into control of a multifunctional prosthesis. Nevertheless, we think that future systems of this kind will be able to achieve better function through the use of position servos based on the principles of Simpson and as adapted by Doubler and Childress for positioning electric-powered joints in space. We also believe that electric-actuated, powered-locking mechanisms will, in the future, ease the effort now involved with locking and unlocking the joints of the body-powered prosthesis with the mechanical levers.
The author believes that provision for natural, subconscious control of multifunctional limbs in meaningful and coordinated ways is one of the great challenges of the medical engineering field. A reasonable medical engineering (human-prosthesis) goal, for persons who require bilateral limbs at the shoulder disarticulation level, is for them to be able to manipulate their environment as well as the best foot users do who have similar arm amputation levels. Of course, if that goal can be achieved, it would mean that we will have also been able to make similar or superior achievements at the less severe amputation levels.
THEORETICAL FOUNDATIONS FOR PROSTHESIS CONTROL
Childress has suggested that general principles for good control seem to come naturally out of observations of the control of various kinds of aids in rehabilitation. He has suggested that these principles may be useful in the formation of a theoretical foundation for control of upper-limb prostheses and has proposed a preliminary theoretical framework. The utility of a theoretical foundation is to give guidelines for control methods so that decisions do not all have to be made only on the basis of experience or subjective feelings. This is one reason for putting forth a set of principles of this kind. Another reason for suggesting an initial set of principles is so that they can then be argued, tested, validated, refuted, altered, modified, or added to. Not all the principles will be discussed in detail here, and the reader is referred to the references for more extended discussion.
The first and perhaps the most important concept, which the author has called Simpsons theory, is based on the following observations: (1) cable-controlled, body-powered arm prostheses-when they can be used-often seem to be controlled well by amputees; (2) Simpson was able to demonstrate good multifunctional control of powered prostheses, without excessive "mental load," by children with high-level bilateral shortage; (3) prostheses that are direct extensions of a limb (e.g., the patellar tendon-bearing [PTB] leg prosthesis) are well controlled; (4) blind people are adept at understanding their physical environment with a long cane; (5) persons with quadriplegia often control their environment well with a mouthstick; and (6) humans in general are very capable when using extensions of their limbs (e.g., stilts, racquets, hand tools, etc.). In all these observations the output is a position variable that is controlled by positions of the body's own joints. These joints, plus the sensation that comes back to the body through the instrument they are operating, seem to provide a natural kind of control that is intuitive and effective. When Simpson implemented this concept for the control of prosthetic arms, he called it extended physiologic proprioception because the physiologic proprioception of the controlling body joints was, in a sense, extended into the prosthesis, much as it is into a tennis racquet or into a hammer when a person uses them. So behind the extended physiologic proprioception concept is the notion that the prosthesis is a kind of "tool" that the body can control very well when it is directly connected in some way to joints of the body. This kind of system has inherent feedback. The output pathway and the input pathway for information flow are both embodied in the tool. This concept for control is the same as the one alluded to in the beginning of this chapter when powered steering of automobiles and cable control of remote manipulators and airplane control surfaces were discussed. Doubler and Childress used tracking studies to provide some objective evidence that this kind of control is superior to "on-ofl" and "proportional" velocity control.
On the basis of these observations and others, on the basis of the objective studies, and on the basis that the cable-operated, body-powered systems as well as Simpson's powered systems were kinds of "existence proofs" of the validity of the approach for upper-limb systems, Childress formulated Simpson's theory of control as follows:
The most natural and most subconscious control of a prosthesis can be achieved through use of the body's own joints as control inputs in which joint position corresponds (always in a one-to-one relationship) to prosthesis position, joint velocity corresponds to prosthesis velocity, and joint force corresponds to prosthesis force.
Carlson, Gow, and others have also worked on this kind of prosthesis control. This control method is relatively simple to implement and has been illustrated in Fig 6D-4.,B. To a great extent this principle suggests that powered prostheses should be controlled in much the same way that body-powered systems are controlled. As with powered steering on a car, required force and excursion can be matched to the force and excursion available by the human operator.
The extended physiologic proprioception control approach realizes feedback of important information in a form that is naturally received by the human operator. This is in contrast to the many kinds of "supplementary sensory feedback" that have been experimented with through the years and that the body does not seem to interpret well. A corollary theory for supplemental sensory feedback, as suggested by Childress, is as follows:
Supplementary feedback can be interpreted best if pressure on the prosthesis is interpreted as pressure on the body (force-to-force correspondence), if the place of stimulus on the prosthesis is represented by a particular place mapping on the body (position-to-position correspondence), and if the velocity of movement of stimulus on the prosthesis corresponds to velocity of stimulus movement on the body (velocity-to-velocity correspondence).
Childress has speculated that direct muscle action can provide the same kind of control that is available from joint position inputs. This direct muscle action control conjecture is formulated as follows:
Natural and subconscious control of a prosthesis can be achieved through the body's own muscles as direct control inputs to position controllers in which muscle position corresponds (in a one-to-one relationship) to prosthesis position, muscle velocity corresponds to prosthesis velocity, and muscle force corresponds to prosthesis force.
The use of tunnel cineplasties (or variants) for control is an example of direct muscle action control. We know that from a control standpoint, these have been successful. New possibilities now exist for expanding the use of this kind of control input, as we have already noted in this chapter. This kind of control may even make it possible in the future to control individual prosthetic fingers in coordinated and meaningful ways. This has always been a hope of many hand amputees.
The myoprehension principle has been described as the natural relationship between muscular contractions and prehension.This is easily illustrated by gripping an object tightly. As the prehension force is increased, muscles of the body that are quite distant from the hand are contracted in reaction to the holding forces being generated. For this reason, it seems natural for the body to relate prehension with muscular effort to some extent regardless of where the muscle is located. Therefore, an EMG signal, which can be related to muscle effort, is a signal that the body can relate to the gripping function. Consequently, the principle suggests that myoelectric control can be somewhat naturally connected with the control of prehension. This is intuitive if the muscles involved are in the forearm but not so obvious if the muscles are proximal to the elbow.
The principles presented can be used as a guide to prescription, provided that the components that are needed are available. We know that good theories tend to fit what is known and can also be used for prediction of new kinds of control schemes. If we apply the principles outlined, they suggest that myoelectric control is good for the transradially amputated limb because the intact elbow, with the prosthesis extension, provides extended physiologic proprioception control of the artificial prehensor in space and the EMG signals independently control prehension (myoprehension). In like manner, for the transhumeral stump, the theory suggests that using a body-powered elbow (extended physiologic proprioception) and a myoelectrically controlled prehensor (myoprehension from the biceps and triceps) is a favorable approach-a fitting principle that has a strong, if not unanimous following as a good solution for this amputation condition. In general, the principles suggest that the body's joints (or muscles) be used as inputs to position controllers (based on the Simpson principle) for control of prosthesis positions (prosthetic joint control) and that myoelectric control, force control through pressure transducers, or possibly a direct muscle input be used for control of prehension. The principles suggest that there are many cases where limbs can be used effectively as rigid extensions of the body, and this implies the need for locking/unlocking of joints. The concepts presented should also be applicable to the lower limb and to other rehabilitation situations where human-machine interactions occur.
As a note of caution, although theories can help guide our decision-making process, in the final analysis they cannot be the final arbiter for prosthesis control decisions, even when they are known to be valid. The final arbiter is the user. Theories have to be subservient to the wishes of the prosthesis wearer and user. The duty of professionals related to the field of prosthetics is to know what is a good (best, if possible) control strategy under given conditions, based on experience or upon theory. If a control strategy based on a theory is in fact good-possibly best-it should also be successful in clinical practice. However, that would only be the case for a large number of fittings. As in statistics, what holds on average may be quite different for a given individual. Theoretical constructs, even when valid, must yield to the will of the individual in deciding the control method finally used, if any. This, of course, does not diminish the usefulness of the theoretical construct unless it happens in a high percentage of cases, in which case the theoretical construct would have to be questioned and re-examined.
Although many factors need to be considered at the time of prosthesis prescription and during subsequent follow-up, the prosthetic control designs that require low conscious control effort by the amputee and that are naturally harmonious in human-machine interactions appear highly desirable and to be the ones that have the greatest potential for minimizing the handicap that may result from a disability due to arm amputation. The theoretical framework that has been presented seems to be congruent with much that we know from previous experience. If it or some modification thereof is valid, it can become an effective guide for prescription. It also appears predictive, which makes it potentially valuable in directing research and development efforts with regard to prosthesis control.
Many of the approaches presented in this chapter do not correspond to the theoretical ideas presented at its end. It was the author's intent to describe a number of the control approaches currently in clinical use and that are commercially available to the practicing clinician. Commercially available systems are not available to implement some of the approaches described in the theoretical construct. Also, as already noted, many complex factors are ultimately involved in prescriptions, with a theoretical framework being only one factor. Also, the framework proposed has been put forward as a "theory" and not as principles that have as yet gained any wide acceptance in the limb prosthetics field.
We have a long way to go before we can say that we have built an "artificial arm" or an "artificial hand." In fact, Beasley and de Bese have said, "There is no such thing as an artificial hand, and the term should be dropped from use as it is misleading." They suggest that "prostheses meet only very specific and limited objectives." By extension of this idea we might say that the upper-limb prostheses currently in use are not worthy of the title "artificial arms." Nevertheless, we can see that progress has and is being made. Powered limbs have perhaps not brought the big advances originally envisioned, but they have taken on significant and practical roles in upper-limb prosthetic procedures, and that demonstrates important progress. We seem to be learning how to integrate them appropriately into practical prosthesis systems.
Progress in science and technology is normally not a linearly increasing function of time. We must continue to seek insights that may result in "breakthroughs" that will yield very rapid improvement of the control of replacement limbs. Short of these "hoped-for breakthroughs," we need to keep making the kind of incremental progress that has brought us to the present state of development.
The author wishes to thank the Veterans Administration Rehabilitation Research and Development Service and the National Institute on Disability and Rehabilitation Research for their sustaining support that has made this paper possible. He would also like to thank his associates Mr. Craig Heckathorne and Mr. Edward Grahn, who assisted and influenced him significantly. In addition, the author wants to acknowledge the clinical assistance of Dr. Yeongchi Wu and Mr. Jack Uellen-dahl, C.P.O., of the Rehabilitation Institute of Chicago, who have been open to use new upper-limb control concepts and who provided a rich clinical environment for this work.
- Abul-Haj CJ, Hogan N: Functional assessment of control systems for cybernetic elbow prostheses-Part I: Description of the technique. IEEE Trans Biomed Eng 1990;37:1025-1036.
- Abul-Haj CJ, Hogan N: Functional assessment of control systems for cybernetic elbow prostheses-Part II: Application of the technique. IEEE Trans Biomed Eng 1990; 37:1037-1047.
- Barry DT, Cole NM: Muscle sounds are emitted at the resonant frequencies of skeletal muscle. IEEE Trans Biomed Eng 1990; 37:525-531.
- Battye CK, Nightingale A, Whillis J: The use of myoelectric currents in the operation of prostheses. J Bone Joint Surg [Br] 1955; 37:506-510.
- Baumgartner R: Moglichkeiten und Grenzen der Prothe-senversorgung der oberen Extremitat. Biomed Technik 1985; 30:340-344.
- Basmajian J, DeLuca C: Muscles Alive. Baltimore, Williams & Wilkins, 1985.
- Beasley RW: The tendon exteriorization cineplasty. In-ter-Clin Info Bull 1966; 5:6-8.
- Beasley RW, de Bese GM: Prostheses for the hand, Evarts CM (ed): in Surgery of the Musculoskeletal System, ed 2, vol 2. New York, Churchill Livingstone Inc, 1990.
- Bejczy AK: Sensors, controls, and man-machine interface for advanced teleoperation. Science 1980; 208:1327-1335.
- Biedermann WG: 1st der Sauerbrucharm noch aktuell? Orthop Technik 1981; 32:156-161.
- Billock J: Upper limb prosthetic management: Hybrid design approaches. Clin Prosthet Orthot 1985; 9:23-25.
- Bottomley AH: Myo-electric control of powered prostheses. J Bone Joint Surg [Br] 1965; 47:411-415.
- Bottomley AH, Kinnier Wilson AB, Nightingale A: Muscle substitutes and myoelectric control. J Br Inst Radio Eng 1963; 26:439-448.
- Carlson LE: Position control of powered prostheses. Proceedings of the 38th Annual Conference of Engineering in Medicine and Biology. Chicago, 1985, p. 48.
- Childress DS: Biological mechanisms as potential sources of feedback and control in prostheses: Possible applications, in Murdoch G, Donovan R (eds): Amputation Surgery ir Lower Limb Prosthetics. New York, Blackwell Scientific Publications, 1988, pp 197-203.
- Childress DS: Closed-loop control in prosthetic systems: Historical perspective. Ann Biomed Eng 1981; 8:293-303.
- Childress DS: Control philosophies for limb prostheses, in Paul JP, et al. (eds): Progress in Bioengineering. Adam Higler, 1989, pp 210-215.
- Childress DS: Historical aspects of powered limb prostheses. Clin Prosthet Orthot 1985; 9:2-13.
- Childress DS: Powered limb prostheses: Their clinical significance. IEEE Trans Biomed Eng 1973; 20:200-207.
- Childress DS, Krick H, Heckathorne CW, et al: Positivelocking components and hybrid fitting concepts for persons with high level bilateral arm amputations. Presented at the 12th Annual RESNA Conference, 1989, pp 296-297.
- Doubler JA, Childress DS: Design & evaluation of a prosthesis control system based on the concept of extended physiological proprioception. J Rehabil Res Dev 1984; 21:19-31.
- Gow GW, Dick TD, Draper ERC, et al: The physiologically appropriate control of an electrically powered hand prosthesis. Presented at ISPO IV World Congress. London, 1983.
- Heckathorne CW, Uellendahl J, Krick H, et al: Achieving complementary manipulative function with bilateral hybrid upper-limb prostheses (abstract). Arch Phys Med Rehabil 1990; 71:773.
- Herberts P, Magnusson R, Kadefors R, et al (eds): The Control of Upper-Extremity Prostheses and Orthoses. Springfield, Ill, Charles C Thomas Publishers, 1974.
- Hoffer JA, Loeb GE: Implantable electrical and mechanical interfaces with nerve and muscle. Ann Biomed Eng 1980; 8:351-360.
- Kovacs GT, Storment CW, Jemes B, et al: Design and implementation of two-dimensional neural interfaces. Presented at the 10th Annual Conferences of the IEEE Engineering in Medicine and Biology Society, New Orleans, 1988, pp 1649-1650.
- Krieghoff R, Bruckner L, Richter KW: Der aktuelle Stel-lenwert Sauerbruch-Prothese bei der Rehabilitation des Armamputierten. Med Orthop Technik 1990; 110:145-147.
- Kuiken T: The Hyper-Reinnervation of Rat Skeletal Muscle (dissertation). Northwestern University, Evanston, Ill, 1989.
- Leal JM, Malone JM: VA/USMC Electric Hand with be-low-elbow cineplasty. Bull Prosthet Res, 1981; 10:52-56.
- Liicke R, Marquardt E, Carstens C: Kineplasty according to Sauerbruch-The fresh indication pectoralis canal in amputations in the region of the shoulder girdle (abstract). Presented at the Sixth World Congress of ISPO, Kobe, Japan, 1989, p 200.
- Marquardt E: The operative treatment of congenital limb malformation-Part I. Prosthet Orthot Int 1980; 4:135-144.
- Michael JW: Upper limb powered components and controls: Current concepts. Clin Prosthet Orthot 1986; 10:66-77.
- Murphy EF: In support of the hook. Clin Prosthet Orthot 1986; 10:78-81.
- Parker PA, Scott RN: Myoelectric control of prostheses. Crit Rev Biomed Eng 1986; 13:283-310.
- Popov B: The bio-electrically controlled prosthesis. J Bone Joint Surg [Br] 1965; 47:421-424.
- Reiter R: Eine neue Elektrokunsthand. Grenzegebiete Med 1948; 1:133-135.
- Ryder RA: Occupational therapy for a patient with a bilateral Krukenberg amputation. Am J Occup Ther 1989; 43:689-691.
- Scott RN: An introduction to myoelectric prostheses, in U.N.B. Monographs on Myoelectric Prostheses. University of New Brunswick Bio-Engineering Institute, Can- 42. ada, 1984.
- Scott RN, Childress DS: A bibliography on myelectric control of prostheses, in U.N.B. Monographs on Myoelec- 43. tric Prostheses. University of New Brunswick Bio-Engineering Institute, Canada, 1989.
- Seamone W, Schmeisser G, Hoshall H: A powered cable drive for prosthetic-orthotic systems, in Gavrilov M, Wilson AB Jr (eds): Proceedings of the Fourth International 44. Symposium on Advances in External Control of Human Extremities. Dubrovnik, Yugoslavia, Yugoslav Committee for Electronics and Automation, 1973, pp 736-755.
- Sears HH, Shaperman J: Proportional myoelectric hand control: An evaluation. Am J Phys Med Rehabil 1991; 70:20-28.
- Sheridan TB, Ferrell WR: Man-Machine Systems: Introductory Control and Decision Models of Human Performance. Cambridge, Mass, MIT Press,1974.
Simpson DC: The choice of control system for the multi-movement prosthesis: Extended physiological proprioception, in Herberts P, et al (eds): The Control of
Upper-Extremity Prostheses and Orthoses. Springfield, Ill, Charles C Thomas Publishers, 1974, pp 146-150.
Thring MW: Robots and Telechirs: Manipulators With Memory; Remote Manipulators; Machine Limb for the Handicapped. Chichester, England, Ellis Horwood Ltd, 1983.
Chapter 6D - Atlas of Limb Prosthetics: Surgical, Prosthetic, and Rehabilitation Principles