Αυτό είναι το “Κεφάλαιο 5” από το βιβλίο “Εφαρμογές Βιοανατροφοδότησης” και έχει τίτλο: “EMG BIOFEEDBACK IN NEUROMUSCULAR REEDUCATION: A PSYCHOPHYSIOLOGICAL APPROACH”. Η Α.Ρ.Α. βιβλιογραφική αναφορά του είναι: “Χρηστίδης, Δ.Α. (2001). Εφαρμογές Βιοανατροφοδότησης. Αθήνα: Έλλην.”
Εάν χρησιμοποιήσετε ως παραπομπή αυτό το άρθρο μου (ή κάποιο άλλο) σε κάποια εργασία σας ή δημοσίευση, σας παρακαλώ πολύ να με ενημερώσετε. Θα χαρώ πολύ εάν μου στείλετε την εργασία σας για ενημέρωση. Σας ευχαριστώ προκαταβολικά.
EMG BIOFEEDBACK IN NEUROMUSCULAR REEDUCATION: A PSYCHOPHYSIOLOGICAL APPROACHΗ
ΗΜΓ ΒΙΟΑΝΑΤΡΟΦΟΔΟΤΗΣΗ ΣΤΗ ΝΕΥΡΟΜΥΪΚΗ ΕΠΑΝΕΚΠΑΙΔΕΥΣΗ: ΨΥΧΟΦΥΣΙΟΛΟΓΙΚΕΣ ΕΦΑΡΜΟΓΕΣ
Το άρθρο αυτό αναφέρεται στη χρήση της ηλεκτρομυογραφικής (ΗΜΓ) ανατροφοδότησης, όπως χρησιμοποιείται στην νευρομυϊκή αποκατάσταση. Αυτή η εξειδικευμένη ψυχοφυσιολογική εφαρμογή απαιτεί καλή γνώση του νευρομυϊκού συστήματος καθώς και των ηλεκτρονικών συσκευών που επιτρέπουν την εφαρμογή μαθησιακών διαδικασιών για την προώθηση της εκπαίδευσης ή/και της αποκατάστασης νευρομυϊκών συμπεριφορών. Επίσης, αναλύεται σύντομα η φύση και τα χαρακτηριστικά του ΗΜΓ σήματος και περιγράφεται το νευρομυϊκό σύστημα με τα χαρακτηριστικά των μηχανισμών του. Στη συνέχεια ακολουθεί συζήτηση των χαρακτηριστικών των νευρομυϊκών παθήσεων για τις οποίες το ΗΜΓ biofeedback προσφέρεται επιτυχώς για θεραπεία και αναφέρονται οι επερχόμενες βιοχημικές και νευρομορφολογικές αλλαγές που σχετίζονται με τη βελτίωση. Επιπλέον, στην τελική συζήτηση προτείνονται ιδέες για μελλοντικές χρήσεις του biofeedback, σε διάφορους τομείς της νευρομυϊκής αποκατάστασης.Abstract
This paper presents the psychophysiological approach of EMG biofeedback, as it is used in neuromuscular reeducation (NMR). This specialized application of biofeedback necessitates a good understanding of the neuromuscular system as well as of the instrumentation that allows the employment of operant learning techniques in the facilitation of motor behaviors. From an organizational point of view, the nature and characteristics of the EMG signal are examined first. Inherent within this examination is a description of the neuromuscular system and its controlling mechanisms. Then, an examination of various instrumentation devices that are used in biofeedback renders the ground rules for understanding the properties of the feedback, as well as the limitations that different instrument configurations might impose on potential applications. The second half of the paper will discuss characteristics of neuromuscular disorders, for which EMG biofeedback has shown to be a successful mode of treatment, and will speculate on the biochemical and neuromorphological plastic changes of the nervous system that affect recovery. Finally, a discussion section will provide suggestions for future applications of EMG biofeedback in different areas of neuromuscular reeducation.
Definition of Biofeedback
The term biofeedback refers to the dual process of monitoring and making a physiological response available for shaping. It involves the use of modern electromechanical devices that quantify an individual’s bodily processes in such a way that they can be fed back to the individual in an immediate, accurate and meaningful way. Neal Miller (1974), the father of biofeedback, defined it as:
[span class=note]“…the use of modern instrumentation to give a person better moment to moment information about a specific physiological process that is under the control of the nervous system but not clearly or accurately perceived. In the terminology of servosystems such information has been called feedback. Such information about a biological process is called biofeedback” (p.684).[/span]
The Neuromuscular System
EMG stands for Electro-Myo-Graphy. An electromyogram is a record of the electrical activity of muscles. With almost no exception, any given human muscle fiber (muscle cell) is innervated by only one alpha motor neuron which is always excitatory (Stein, 1980). An alpha motor neuron, whose axon originates in the ventral horn of the spinal cord (or in some cases in a cranial nerve motor nucleus), generally branches to excite a small group of muscle fibers, rather than only one single fiber. The motor neuron together with the group of muscle fibers that it innervates, is called a motor unit (see figure 1). A motor unit is the lowest level of hierarchical organization in the vertebrate nervous system, providing the only route by which the central nervous system can influence the contraction of a particular muscle fiber. For this reason, the motor neuron is often referred to as the “final common path” of the nervous system (Sherrington, 1906). Within the motor unit, a single action potential in the motor neuron evokes a large excitatory junctional potential, followed by an action potential and then a twitch in each of its muscle fibers. Larger contractions of these muscle fibers are the result of a rapid succession of action potentials (rate coding), such that each twitch begins before the previous twitch has fully terminated. These electrical events that are triggered by the arriving spike of the motor neuron are called motor unit potentials (MUP’s) or myopotentials. As additional motor units are being activated, the muscle as a whole begins to contract and this process is called recruitment. The summation of MUP’s that are triggered by the spiking motor neurons provide the electrical activity for EMG.
Reciprocal inhibition and reciprocal innervation
Motor neurons within the same muscle, as well as those of neighboring muscles, are usually excited together by a determined group of presynaptic neurons. Such a collection of motorneurons that are recruited in synchrony to produce a common action is designated as a motor neuron pool. The co-working muscles that such a motor unit pool innervates are called synergists. Muscles that work in mechanical opposition to these muscles are regarded their antagonists. When a synergistic muscle group is contracting, their antagonists remain relaxed as a result of the central nervous system mediated inhibition that is called reciprocal inhibition. Reciprocal innervation is another type of inhibitory mechanism involved in the control of movement. It is triggered by the stretch reflex and mediated by direct monosynaptic connections onto inhibitory neurons within the spinal cord. Both forms of inhibition rely on afferent input from the I-a primary spindle afferents, and the inhibitory neurons that mediate the response are the I-a inhibitory neurons. The essential difference between the two forms of inhibition is that reciprocal inhibition can be influenced by descending control from higher centers via ventral input onto the gamma motor neurons, whereas reciprocal innervation, as a reflex, is not subject to influences from higher centers (see figure 2).
Function of spinal circuits in reciprocal innervation. Reciprocal inhibition is an important CNS mechanism for the control of movement. As shown in figure 2a, a branch from an extensor spindle efferent (I-a primary) synapses on a I-a inhibitory interneuron which in turn releases an inhibitory transmitter substance at its synapse with the flexor motoneuron. We can appreciate their importance by recalling Dale`s law: “A mature neuron makes use of the same transmitter substance at all of its synapses” (Eccles, 1957). As it follows from Dale`s law, it would not be possible to get reciprocal inhibition by allowing a branch of the extensor efferent to synapse directly on the flexor motor neuron. This would produce a non-functional state of reciprocal facilitation, where both the agonist and the antagonist would be firing simultaneously and exerting antagonistic forces on the same joint.
Recurrent Inhibition. A more complex form of inhibition is mediated by the Renshaw cells in the spinal cord and it is called recurrent inhibition. As figure 3 shows, the Renshaw cell receives direct input from collateral branches of (alpha) spinal motor neurons (which have been activated by monosynaptic excitation from I-a afferent input from the spindles) and, in turn, inhibits many motor neurons, including the one that gave rise to its input (Carew, 1983; Schwartz, 1983). The Renshaw cell has many functions but the most important one, is probably that of limiting the motor output interval of the particular motor pool that gave rise to the stimulus that it received from the alpha motor neuron collateral. Since it activates the I-a inhibitory interneuron, at the same time that it inhibits the homonymous and synergistic motor neurons, it disinhibits the antagonist by inhibiting the I-a inhibitory interneuron. Thus, the Renshaw cell is in a position to influence the magnitude and duration of the I-a afferent-mediated reflex response that can not be otherwise influenced by the central nervous system.
Stretch-reflex co-contractions. Reciprocal and recurrent inhibition assure that antagonistic muscle groups do not oppose each other mechanically. An exception to this rule occurs during very fast whip-like motions of hinge-type joints. There, the work of antagonistic muscles serves as a protective mechanism against damage from the strong forces that the prime mover could produce (Barnett and Harding, 1955; Basmajian 1957, 1959; Bierman and Ralston, 1965; Guyton, 1981; Henneman, 1974). There is strong consensus among investigators, that stretch-reflex co contractions among antagonists do not occur during slow movements (Angel, 1975; Guyton, 1981; Hallet et al., 1975; Henneman, 1974; Morin et al., 1976; Patton and Mortensen, 1971). Research by Lundberg and his coworkers (1975) shows that prevention of co contractive patterns is accomplished though II and I-b afferent fiber excitation (from the spindle nuclear chain fibers and the Golgi tendon organs, respectively) of polysynaptic pathways in the cord, and that these pathways are subject to intervention from rostral centers.
The properties of synergism and reciprocal inhibition can be used advantageously, when providing EMG biofeedback, to facilitate the flow of the neuromuscular signal towards a certain muscle, to inhibit others, and most important, to determine the patient’s position on the continuum of treatment considerations which will be discussed below. As it will be pointed out in the section on clinical applications, these properties are overlooked by the great majority of biofeedback clinicians.
Properties of motor units
Size and firing order: Since facilitation of motor unit recruitment is an essential part of EMG biofeedback, it is important to become aware of the functional significance of the properties of motor units. It might seem that the CNS can recruit for use any motor unit out of its available supply, according to the needs of the moment. This is not the case. The firing of motor neurons is related to their dimensions, which, indeed, extend over a wide range. As established by the size principle, the motor units of a muscle are fired in a selective order, which is determined by the sizes of the motor neurons. The smaller the motor neuron, the easier it can be fired; the larger it is, the greater its threshold to depolarize. Since the size of a motor neuron and the size of a motor unit are directly related, it follows that the participation of a motor unit in a motor unit pool is dictated by its size (Wuerker et al, 1965).
Workload and recruitment: There is by far a much greater number of small motor units in a muscle than there are large ones. Each has a slightly different susceptibility to discharge which is, of course, proportional to its relative size. Small contractions are induced and precisely controlled by discerning activation of varying numbers of these small motor units. When the total workload of the muscle needs to be increased, larger motor units, providing larger increments of tension are recruited. The experimental evidence indicates that larger units never become active without the total recruitment of the smaller units, due to the fact that this order of activity is fixed centrally (Guyton, 1981; Henneman and Olson, 1965; Wuerker et al, 1965). Another principle of recruitment is that, as the need for workload increases, fewer and fewer additional motor neurons need to be recruited. As for the application of these principles to biofeedback, the thing to keep in mind is that in order to recruit the full muscle`s capacity of motor unit firing it is necessary to a) be able to accurately measure and provide feedback of the increasing summation of MUP’s, and b) to provide increasing workload demands on the muscle in order to facilitate recruitment of successively larger motor units.
Voluntary vs. reflexogenous motor behaviors.
The utmost goal of EMG biofeedback in neuromuscular re-education is the establishment of voluntary and functional neuromotor behavioral patterns. Voluntary neuromotor behavioral patterns differ from involuntary and stereotyped reflexive behavioral patterns in the sense that one does not have to learn and practice how to breathe, swallow, or to withdraw his/her hand from a hot stove. These and many others, are highly stereotyped reflexive motor behaviors, where no aspect of an individual’s attention can change their essential character. On the contrary, we need specific training in order to produce voluntary motor behaviors such as speech, handwriting, playing a video game, throwing a curve-ball, and numerous other behaviors that range from the inconsequential to the extraordinary. These acts are highly variable, very frequently idiosyncratic, and vary notably among individuals.
The Keele-Summers model. The process of acquiring a new neuromotor behavior is best described by the conceptional model developed by Keele and Summers (1976). The Keele-Summers model suggests that any skilled movement generates various types of information that are of extreme importance for the acquisition of that specific motor skill. Neuromuscular input is provided to the muscles involved by a hypothetical movement program. Feedback from receptors in the joints and the muscles, as well as other types of information regarding movements (visual, vestibular, proprioceptive, etc) are matched and compared to a model of skilled performance. Feedback information constantly furnishes detail about errors and successes (hits or misses) in such a way, that if the behavior is rehearsed sufficiently the errors are minimized or totally eliminated.
Engrams. In the final stage of neuromotor behavior acquisition, the skilled behaviors become automatic. Welford (1974) notes that at this level of neuromotor skill learning, the conscious awareness of the act dissipates and the act is realized by automatic execution. Faster execution of neuromotor behaviors becomes feasible because feedback control loops grow less important as the decision making of a motor act becomes automatic or neurally coded. Many investigators refer to these hypothetical, localized memory traces that are the initiators of preprogrammed motor actions as engrams (Lansley, 1950; Reinis & Goldman, 1982; Taub et al., 1975; Thompson, 1980). Even though the search for an engram parallels the process of a Sisyphean act, the term will be employed here symbolically, simply because of its heuristic value in describing closed-loop and open-loop control mechanisms as mediating components of neurally coded motor behaviors.
Closed-loop control systems
A closed-loop control system is one that communicates information ceaselessly from a system (or device) that is being controlled to the device that controls it. Its principal components are a transducer, an error detector, and a controller (see figure 4).
The transducer receives feedback from the output of the controlled device. The error detector receives input from the transducer and a) compares the actual output to the desired output, and b) generates a new “action command” based on the feedback that it received from the error detector. Finally, the controller receives the new and upgraded “action command” from the error detector and activates the controlled device (Houk, 1974; Houk and Henneman, 1974; Rosenzweig and Leiman, 1982).
An example. Closed-loop controlled systems range from simple, such as a thermostat that controls room temperature by activating a furnace and an air-condition as needed, to complex such as the human endocrine and circulatory systems. A good example to elucidate the function of a closed-loop system in neuromotor behaviors is that of a bicycle rider. The transducer can be likened to the rider’s visual system, the error detector to some properties of the perceptual, proprioceptive, and kinesthetic system, and the controller to the neuromuscular system that is used for steering. Through the interaction of these systems, the actual position of the bike/rider combination on the road is continuously contrasted with the desired position, and corrective commands from the controller are fed to the muscles that control steering for appropriate action. It follows that a closed-loop system allows for moment to moment corrective changes based on information that is being received from peripheral sources.
Open-loop control systems
Open-loop control systems are those that do not depend on external sources of feedback (Houk, 1974; Houk and Henneman, 1974; Rosenzweig and Leiman, 1982). Open-loop systems are extremely important in the execution of rapid preprogrammed actions where there is no time wasted for the processing of external feedback. Open-loop systems are being manipulated and monitored by external control and sensor mechanisms, but their activity is preprogrammed. Open-loop control systems are frequently confined within closed-loop control systems, and the neuromuscular system makes frequent use of this arrangement. By accounting for potential errors in the preprogramming of an open-loop control system, in addition to offering high speed of response, the system also guarantees freedom from variability and error.
An example. An assembly line robot would offer an excellent example of such a system in the world of electromechanical devices. In the world of neuromuscular systems, the bicycle rider example can be used afresh. Riding a bicycle is not as elementary as it might seem, even though “any body can do it”. If the operation of each component part had to be analyzed and examined individually, it would soon be realized that if each component part had to be processed and executed discretely as well as in coordination with the rest, one would never be able to even get his/her feet off the ground. This is to say, bike riding is an activity that can not be taught. Nobody can actually learn from a lecture and a set of instructions how to produce and coordinate (bilateral) activities like dorsiflexion, plantarflexion, knee extension and flexion, forearm extension, flexion, rotation, and pronation, wrist and finger extension, flexion, rotation, pronation, while trying to keep his/her balance and an eye on the road. Realizing that the motor actions just mentioned are just a few from what it actually takes to ride a bicycle, the importance of open-loop control systems is to be appreciated. Based on the ideas of Keele and Summers (1976) and Welford (1974) bicycle riding is learned by “trial and error” and perfection can be achieved only through continuous rehearsal. Both closed-loop and open-loop control mechanisms lead to the formation of engrams that make fast and complex motor acts possible with a minimum amount of feedback.
A combined example
Increasing familiarity with the operational characteristics of computers and automation devices has led to an increased tendency to implement them in examples that attempt to describe human brain mechanisms and interactions with the periphery. Here, an assembly line robot is used in an effort to explain what can go wrong during the various relay stations of the neuromuscular command.
The engram. The program that “drives” the robot’s mechanism to execute a particular routine is written and stored in the robot’s memory. The creation of that program goes through various stages of development. By methodologically observing the robot’s performance and comparing it to the desired performance, the programmer keeps upgrading and correcting the program’s routines to perfection. At the point where the program is running smoothly and without errors, the programmer stores it in the robot`s memory along with other programs that were created in the same manner. Whenever the programmer wants the robot to perform a certain task, all he has to do is to feed the program that corresponds to that task into the robot`s controller mechanism. This process then, is similar to that of creating an engram in the human equivalent.
Problems at the top. The faithful robot will keep performing its tasks until one of its memory chips blows out. The controller mechanism as well as its peripheral mechanism remain intact, but the messages that are fed into them (if any), are incomprehensible and the robot (or parts of the robot, depending on how significant the chip was) either stops working, or malfunctions. This would be the human equivalent of a cerebrovascular accident (CVA) (or some form of closed head injury) where the periphery is intact, but brain tissue has been destroyed or injured as a result of the vascular lesion and the concomitant built up of local pressure. If the blown chip is in the controller, then the robot is left with a good program of how to run but the instructions can not be relayed accurately from the memory to its peripheral electromechanical devices. This condition in humans can resemble that of a stroke that is localized specifically within motor relay areas and interferes to various extends with transmission of neuromotor commands. The same condition can also resemble cerebral palsy, where the brain injury or insult interferes with early brain growth. In both cases, ballistic movements, ataxia, and spasticity are the observed symptoms.
Problems at the end. In another situation where the programmer gets upset with his working conditions and displaces his anger by pulling off the robot`s main cable, the robot remains with a perfect engram, a functional controller that attempts to transcribe and distribute the various commands to the peripheral devices, and the peripheral devices remain intact as well. Then, the amount of commands that the periphery will receive from the controller becomes directly disproportional to the amount of wires that the raging programmer managed to yank from the main cable. As it follows, the peripheral devices that will become incapacitated will be those whose wires have been snapped-off. This condition in humans resembles a spinal cord injury where direct forces on the cord or compression by bone fractures (as the result of forceful acceleration, gun shot wound, or other accidents), can injure and destroy neural tissue. Loss of function is directly related to the degree and level of the injury. In the next section, rather than having more time spend on creating parallels between human conditions of motor loss and incapacitated robots, the role of EMG biofeedback feedback in the restoration of those conditions will be examined.
A trial and error approach.
What is called “repair” in mechanics and electronics, is called “rehabilitation” in physical medicine. If someone were to repair the robot after the departure of the upset programmer, he would have two alternatives: to either follow the blue prints and reconnect the wires as they were originally coupled in the main cable, or to proceed by trial and error. Going by trial and error is the only alternative in rehabilitating physical disease -since the “engineer” did not make the blue prints available. In the case of the robot, the repairman has to try all possible wire combinations, and based on the feedback readings that he gets on his oscilloscope or whatever equipment he uses, he would know whether he is making the correct connection or not. Similarly, EMG biofeedback equipment supply both the patient and the therapist with information of whether a specific response triggers the desired neuromotor activity or not. Based on information derived from continuous trial and error attempts, a new neuromotor behavior program can be initiated, or activation patterns of new motor units can be triggered from neighboring motor neuron pools. In both cases, rehearsal with correct feedback can lead to elimination of errors and establishment of new functional engrams.
Evolution through learning: A working model
The basic assumption that is being developed so far is that EMG biofeedback facilitates neuromuscular reeducation by providing essential information to the brain since the brain has lost the ability of acquiring that information on its own. At what point then, within the system, is this information obtained, and to what open-loop control system(s) does it contribute? As much as our knowledge of motor behaviors has been enhanced during the recent decades, this question still remains unanswered (Talbot & Humphrey, 1979). To date, we are still unable to understand the complete Neurophysiology of even the simplest neuromuscular command! Neuromuscular behaviors have been studied by neurobiologists, physiologists, neuroethologists, zoologists, bioengineers, and orthopedists (natural historians, anatomists, sports scientists, dancers, etc, should probably be included in the list as well). The beauty of this great diversity is that by bringing in their own prejudices and windows for looking at the world of motor behaviors, they also bring along their great talents. The field by each own nature is broad enough to, not only accept this great diversity, but to welcome it as well for the creative freshness that it brings along. Excellent (and beautiful) books have been written in an attempt to grasp the essence of motor movement by the physicist Borelli (De motu animalium, 1680), the engineer Hertel (Structure, Form and Movement, 1963), the zoologists Gray (Animal Locomotion, 1968) and Alexander (Animal Mechanics, 1968), the neurobiologist Roberts (Neurophysiology of Postural Mechanisms, 1978), and the bioengineer McMahon (Muscles, Reflexes, and Locomotion, 1984).
Through the contributions of scientists, such as the ones just mentioned, we do have enough information to build a working model that can help our understanding of EMG biofeedback treatment, regardless of whether it reflects the true state of events or not. The model is adopted from the principles of the Neocortical Override Theory (Allen, 1983), and its main function, as presented here, is to make the reader aware of the hierarchical dominance that exists within the neuromuscular system as well as of the unquestionable importance of subordinate levels. Adopting a synthetic perspective, and since learning is the underlying principle of EMG biofeedback, we will assume that learning has played an important role in the evolution of the neuromuscular system. In other words, the fittest have been the ones who were able to learn, and that organisms are capable of learning. Faster rate of learning permitted them to be the fittest in their environment and therefore survive. The subject matter of this paper does not permit further discussion of this premise, but the reader should accept it as a cross-breed of Lamarckian and Darwinian dogma where learning processes have been reinforced by the organism’s environment and thus contributed to the ultimate development of the neuromuscular system. The biochemical and neurophysiological evidence relating to the amazing degree of plasticity that exists within the nervous system, is discussed later in the paper. The spontaneity by which the nervous system launches and regulates these plastic changes, should make one wonder whether a learning dimension has not played, indeed, an important role in evolution.
Spinal cord: The simplest and most primitive (and, by implication, vital) neuromotor behavior is that of locomotion. Spinal reflexes, the simplest components of a neuromotor behavior predominate. Fish, for example, began to move forward by learning to send waves of bending backward down their spinal column. This is accomplished by periodically contracting the muscles on one side of the body while the muscles on the other side remain silent. Does the principle of reciprocal inhibition come to mind? In order for this wave of alternating bilateral contractions to travel down the spine, a phase lag between rostral and caudal segments was developed. Using the principles of rate-coding and recruitment, fish are able to move faster by increasing the frequency of the tail-beat motions (Grillner, 1975; Hertel, 1963).
Brain stem: As time went by and the crossopterygia were demandingly attempting to conquer the dry land, their pelvic and pectoral fins begun to transform into the forelimbs and hindlimbs of the primitive tetrapods. It is exactly within this transformation process that learning played an important role in the evolution of neuromuscular systems: the faster an organism could learn to move faster, the better its chance of conquering a new environment first, and the better its chance of surviving and continuing to evolve. On land, primitive tetrapods begun learning that rather than undulating, they were better off extending their limps away from the body and discovering some manner of walking. This marked a rather interesting period in the evolution of spinally controlled neuromotor systems: Neural circuits that became responsible for pattern generation and the subsequently developed bipedal and quadrupedal gaits emerged.
Somewhere along the same lines, these primitive organisms realized that the further they could sense their surrounding environment, the better their chances of surviving within that environment. Thus, learning to enhance their sensitivity to certain body parts, that themselves were more sensitive to stimulation from their surrounding environment (probably due to physical proximity with the stimulus), was reinforced and this marked the evolution of the senses. Since direct contact with the physical surroundings characterized this time period, sensory organs within the muscles were developed (Golgi tendon organs and muscle spindles) that permitted the organism to navigate reflexively through the physical boundaries of its environment.
The evolution of both the neuromuscular system and the senses imposed new demands on the capacity of the nervous system. As a more complex environment became available to the evolving organism, the nervous system broke away from the confines of the spinal cord (which by then, was itself evolving into a quite complex structure), and the brainstem developed. Its basic structures: the medulla oblongata, the mesencephalon, and the pons emerged from the intricate circuitry whose ascending information allowed them to monitor and regulate interactions with the external, as well as the internal environment. Consequently, this greater capacity of the mesencephalon to receive and assimilate efferent information, allowed for even more complex forms of interaction with the periphery and the external world.
Hypothalamus, pituitary, and cerebellum: As more intricate systems developed for the musculature and for the senses, the boundaries of the external environment extended even further and the organism began to interact with it in even more complex ways. The organism acquired a need to adopt to the external world’s more demanding and acute circumstances. A new, higher order of structure level developed atop of the mesencephalon. The basic structures of this level are the hypothalamus and the pituitary gland that regulate internal homeostasis as the organism responds to the needs of its environment, and the cerebellum that allows for coordination and smoothness of motor functions.
Limbic system: A higher level of organization was fashioned along the same lines: the limbic system (paleocortex). The limbic system probably owes its origination to certain neural regions that became increasingly sensitive to information relayed to it by the senses. The organism learned that certain types of information always called for the same type of physiological reaction. Thus, exposure to a certain kind of stimulus kept leading to a stereotyped response which, according to Welford’s ideas (1974), became neurally coded. Further increases in sensitivity, through the evolutionary process, reinforced the growth of these structures. The basic function of the limbic system, at that time in evolution, was that of triggering alterations in the homeostatic regulation of the subordinate level. Thus, the “fight-or-flight response” entered the scene which permitted the system to modify bodily activity in response to such triggers as fear and anger, by overriding the visceral regulating activity of the subordinate brain structures.
Thalamus: The neuronal circuitry that allowed for reciprocal connections between body structures and the CNS and the vast connectivity within the brain and the spinal cord created a need for a “master switch-board”. This master switch-board is none other than the thalamus. It could be argued that the thalamus has the role of the “master controller” in the highest order closed loop control system in the brain.
Cerebral cortex: From an evolutionary perspective, the cerebral cortex (neocortex) is the newest of all of the brain’s structures. This area that is associated with high order thought processes, including memory, intellect, logic, intuition, interpretation, language, cognition, and other facets of information storage and processing, also functions as both the transducer and the error detector of the highest order control loop of the neuromuscular system. Motor commands flow endlessly through the “master controller” thalamus to loops that connect with the periphery.
A man on a horse that rides the “horse”. Thomas McMahon (1984) in his book: Muscles, Reflexes, and Locomotion presents the developmental history of walking robots and vehicles with legs that begun in preparation for the exploration of the moon and perhaps other planets by the United States and the Soviet Union in the 1960’s and 1970’s. The development of these machines followed a strikingly similar pattern to that of the evolution of the neuromuscular system but without intending to copy its characteristics. The pattern of similarities started to develop when some of these machines started to employ force and position feedback from the legs, and a few even had simple reflexes that prevented the legs from tripping over obstacles or stepping in holes. The first vehicle with legs that was able to walk by itself under computer control was the “Phoney Pony” (McGhee, 1966; 1968) which had four legs and was powered by electric motors. Two twelve volt car batteries provided the power, and each joint had one degree of freedom. It weighted about fifty kilograms and had a top speed of 0.5 miles per hour. Even though it had the size of a small pony, nobody could ride on it. A walking vehicle suitable for being ridden would have to be much larger.
Two engineers at the General Electric Laboratories (Liston & Mosher, 1968) developed their own version of a “Phoney Pony” whose dimensions could easily put it in the “Trojan Horse” category. Their vehicle weighted about 1.5 metric tons and a 100 h.p. motor supplied the hydraulic power via trailing hydraulic lines. The rider’s body had to be strapped into a seat, and he could control the twelve massive joints through a combination of levers. He could lift one of the rear limbs of the machine by lifting his foot that was strapped into a harness. Each of the four legs had three degrees of freedom, so by extending or flexing his knee, or by abducting and adducing or rotating medially and laterally at the hip he could induce the same movements in the machine’s legs (the machine did not allow for dorsiflexion, plantarflexion, or medial and lateral rotation of its “foot”). The front limps of the machine were controlled in a similar fashion with hand controls. In addition, force feedback from the machine’s legs could alter the “feel” of the levers, so that the rider could know what the legs were doing. Even so, manipulating the machine’s twelve joints was so physically demanding and required so much concentration that most riders could do it for only a few minutes at a time. The major problem seemed to be that human riders are not accustomed to controlling and keeping track of four legs.
In the midst of all this frustration, someone suggested, probably facetiously, that… a horse should rather be operating the vehicle: by strapping the horse into something like a suit of armor with potentiometers attached in the joints, then a person could ride the horse and the vehicle would follow the horse’s movements. Well, even though this suggestion might have turned out not to be such a bad idea, it was never tried out. Had this “Trojan Horse” venture ever reached a functional state, the rider could play the role of the sensorimotor cortex and the brainstem, while the horse would be the pattern generators of the spinal cord. In order for the vehicle’s legs to move, both the rider and the horse would have to coordinate their own motor output in a way that it would produce a functional result. Finally, faulty or misinterpreted signals, regardless of their origin (the rider or the horse), would produce a non-desirable movement in the vehicle’s legs.
In the event that the curious reader is wondering what is the current status of the legged-vehicles research, here is a piece of “iron horse” trivia: The Adoptive Suspension Vehicle was successfully tested in the fall of 1985. Sixteen computers are coordinating the movement of its six legs, and sixty Ohio State engineers worked intensely on its development for the past four years. It has a top speed of eight miles per hour, and it can step gingerly over terrain where wheels can’t roll. Ohio State engineer Kenneth Waldron was quoted saying: “Its like a mechanical mule…” (Hammold, 1985).
In the beginning of this section, the EMG biofeedback instrumentation was paralleled with the repairing equipment that a technician would have used in order to inspect the integrity of the circuits in the malfunctioning robot. In a manner that is very similar to the way that an oscilloscope allows a repairman to appraise the integrity of a connection as well as the exact parameters of the voltage drops across a point in question, the biofeedback instrumentation allows the clinician to “tune into” a muscle of interest and to evaluate the integrity of the flow of the neuromuscular command that descends from the cortical areas through the cord and terminates at the muscle.
It is important to be understood that it is only the characteristics of the descending neuromuscular command that can be assessed by the biofeedback instrumentation. Lesions or other problems in the ascending neuromotor and sensorimotor pathways are evaluated with the diagnostic skills of a neurologist and\or the use of radiographic contrast procedures, (such as myelography that permits the assessment of compressive lesions or those that distort the cord from within), selective catheterization of spinal blood vessels and injection of radiopaque dyes (such as spinal angiography that permits the assessment of vascular lesions), CAT-scans, MRI’s and other diagnostic devices.
The biofeedback instrumentation, without taking into direct account the site of the lesion, permits commencement of a standard treatment protocol contiguous on the evaluation’s findings. This simply means that the clinician, having completed his/her evaluation of the patient’s condition, will treat the patient without being excessively preoccupied with the exactness of the patient’s diagnosis. It is as if the “trojan horse” rider would be looking at the legs of the vehicle and receive constant feedback as to whether his command produced the desired movement or not. In the event that it did, he would try to remember it so that any time he wishes to produce the same movement again, he can repeat it. Without doubt, rehearsal would increase the probability of its recurrence. In the event that his command did not produce the desired movement in the vehicle’s legs, he would go through a trial-and-error routine of giving the horse particular commands and observing the resultant movement in the vehicle’s legs. When he would finally see the desired movement, he would try to repeat it by recalling what his last command was. Given that the horse in this ideal example always omits the same motor behavior in response to the particular command that it receives, recalling the latest command will again produce the desired movement in the vehicle’s leg, and the response can be rehearsed over and over until it becomes dominant in the rider’s command repertoire. In a similar fashion the EMG biofeedback instrumentation plays the role of the rider’s visual system and relays information of whether a given response occurred or not. Rather than looking at the produced movement however, it looks at the response of the muscles that produce that movement. The following section will describe and explain the components of the EMG biofeedback instrumentation.
EMG biofeedback instrumentation
Familiarization with the components, characteristics, and properties of the ideal EMG biofeedback instrumentation is necessary in order to appreciate and make correct use of the biofeedback principle. The basic EMG biofeedback instrumentation consists of the following components: electrodes or sensors, differential amplifiers, filters, rectifiers, integrators, level detectors, and displays. The role of each of these will be explained briefly below. An in depth analysis is beyond the scope of this paper and the interested reader is advised to look up Basmajian (1983, 1967), Cohen (1983), Peffer (1983), Paskewitz (1983), Iscoe and Young (1981), Stern, Ray, and Davis (1980), Cuha and Anand (1979), Kwatny, Thomas, Gottlieb and Agarwal (1970), and Kwatny (1970) for more technical details.
The special sensors that are able to “translate” ionic potentials into electric potentials are called electrodes. All the different varieties of electrodes that are used in the neurosciences can be easily categorized into three basic groups: micropipettes or microelectrodes, indwelling or needle electrodes, and surface or bipolar electrodes. Of the three types only the needle and the surface ones are used for the measurement of biopotentials such as the EMG. The measurement of EMG requires two electrodes since the measured voltage is essentially the detected difference between the potentials of the two electrodes. Since most EMG biofeedback equipment are using differential amplifiers (see bellow), a third electrode (ground) becomes necessary.
Needle-indwelling electrodes: Bipolar fine-wire “needle” electrodes are introduced transcutaneously directly within the muscle mass, exactly at the site of interest. Their manufacturing process has undergone such a significant evolution within the past decades that their use is no longer as forbidding as it used to be (Basmajian, 1972). They are made from a nylon (teflon or polyutherane) isolated alloy (e.g. nichrome, platinum, stainless steel) that measures only 25 to 75 microns in diameter. Their small size allows for painless insertion and withdrawal. Due to the fact that they can be inserted directly into the muscle of interest they have the advantages of a) making deep muscles available, b) being able to localize specific muscle sites and reduce “cross talk” between muscles, and c) providing increased sensitivity when monitoring single motor-unit activity (Basmajian and Blumenstein, 1983). Despite their advantages they are rarely used in EMG biofeedback mainly due to their invasive nature, demand for sterility, and lack of experience and/or medical qualification of the average biofeedback clinician.
Surface silver-silver/chloride electrodes: The use of surface electrodes makes EMG biofeedback non-invasive, convenient and approachable to clinicians from various backgrounds. There are very few reports in the literature that compare EMGs recorded with needle and surface electrodes from the same sites. These limited reports show that under certain provisions the results are similar (Basmajian, 1976; Komi and Burkirk, 1970). Professor Basmajian, being the author of more than 200 scientific articles and textbooks, as well as one of the founders of the International Society of Electromyographic Kinesiology in 1965 (now called the International Society of Electrophysiologic Kinesiology) is an undisputable authority in electromyography. Throughout his writings he encourages clinicians to either use needle electrodes, or to make sure that they are using quality surface electrodes, to apply them correctly, and to understand the limitation of monitoring superficial muscles where specificity becomes a function of the relative placement (Basmajian, 1972, 1974, 1976, 1983).
There are many different kinds of surface EMG electrodes. Lately the market has been saturated with ready-to-use disposable silver/silver-chloride electrodes that minimize application time and they are relatively inexpensive. Silver/silver-chloride electrodes are characterized by their ability of resisting polarization and eliminating noise (Cohen, 1983). As a rule of thumb, the ideal type of EMG electrode is the one specified by the manufacturer of the equipment (Peffer, 1983). As it will be described in the “amplification” section, a good match between the electrodes and the input of the amplifier is essential. An EMG biofeedback electrode configuration consists of two reference and one ground electrode. The reference electrodes are also called “actives”, and the ground is sometimes called “inactive”. The reference electrodes are placed in a bipolar configuration along the vector of the muscle, and the recorded EMG is the algebraic sum of the action potentials that are produced from the contracting muscle. Such an electrode placement will record the difference in electrical potentials that originate between them, and to a lesser degree those that originate in surrounding and distant muscle tissue.
Placement. The placement of the pair, as well as the inter-electrode distance determine which muscles will contribute to the reading. Closely spaced electrodes (about one inch or less apart), make better provisions for observing single motor units. Electrodes that are placed at equal distances but further apart from the belly of the muscle (depending on the size of the specific muscle), provide a good index of overall muscle activity (Basmajian, 1983; Guha and Anand, 1979; Stern, Ray, and Davis, 1980).
Skin preparation. Skin preparation procedures for electrode placement should be considered seriously. Proper skin preparation and electrode placement can bring the resistance of the electrode-to-skin contact to as low as 3000 ohms (Basmajian, 1974; Fernando, 1983; Peffer, 1983). The following are simple guidelines for proper electrode placement:
- Cells in the stratum corneum (utmost layer) of the epidermis, having undergone degenerative changes, form a layer of dead, flattened cells which is very rich in keratin (Shepherd, 1983). This layer of dead cells is not only resistant to water, most chemicals and enzymatic digestion, but is also resistant to electrical conductivity. It is important therefore to scrub and remove the stratum corneum as well as the naturally produced lipids and body oils in order to increase the electrical conductivity of the skin and reduce the electrode-to-skin resistance.
- One good way of reducing movement artifacts from the electrode-skin connection is to use the cup-type surface EMG electrodes. Good quality cup-type electrodes have found wide use in electromyographic applications as a result of their reliability (Basmajian, 1974; Goodgold and Eberstein, 1972). The part of the electrode that comes in conduct with the skin consists of a cup-shaped plastic housing whose bottom surface has been coated with a silver/silver-chloride solution. The inside of the cup is filled with an electroconductive “jelly” or “paste” that acts as a bridge between the skin and the silver/silver-chloride coated surface of the electrode. Since the jelly mass responds readily to movement, it ensures that the electrode-to-skin conduct is not interrupted.
- There should always be plenty of jelly in the cup to assure good contact. Air spaces will produce artifacts. When adhesive collars are used to secure the electrode placement, pressure should be applied to the cup first. This ensures a good contact in the skin\electrode\sensor configuration, and at the mean time prohibits the formation of air bubbles in the cup. On the other hand if too much jelly is being used, as it is being squeezed outwardly by the applied pressure, chances are that it will interfere with the collar`s adhesive properties, and it might even short the two electrodes out if the inter-electrode distance has been smothered with jelly.
- It is always a good idea to check the skin-electrode connection after the electrodes have been attached. Both an impedance meter or an ohm meter will do the job, but an impedance meter is to be preferred. The ohmmeter introduces a small D.C. voltage that can polarize the electrodes, whereas the impedance meter introduces a small A.C. current that eliminates the polarization problem.
The EMG signal that is picked up and travels through the electrodes is a very weak bipolar signal whose amplitude ranges typically between 1-250 uV, (although amplitudes reaching 600 uV are commonly recorded during high resistance heavy muscle work loads) and has a frequency that ranges from 20-10,000 Hz but peaks between 30-100 Hz. Frequencies above 200 Hz contribute minimally to the total voltage (Grossman & Weiner, 1966; Kwatny, Thomas, & Kwatny, 1970; Lynn et al, 1978) and therefore can be ignored. The function of the amplifier is to magnify the weak bi-polar current that is picked up by the electrodes and to bring it within a range that is compatible of driving (and/or being processed by) other devices. The amount of amplification that it delivers is called gain and it is stated as the ratio of the output to its input.
A certain type of amplifier named differential amplifier is the one most commonly used in EMG biofeedback instrumentations, due to its ability to reject noise. What makes a differential amplifier unique, is its ability of rejecting signals that are common to both electrodes. To understand its operation we can think of it as a combination of two amplifiers with a common output but separate inputs. The output remits the amplified sum of the two inputs. This operational principle necessitates the use of the ground electrode that was mentioned previously, such that each input receives signal from one active electrode “in reference to a common ground”. Both amplifiers have the same gain but while one is inverting the signal, the other one is non inverting. If both inputs of a differential amplifier are connected to the same signal source, the output of the device should be zero since it will be summing the same inverted and non-inverted values. By the same token, noise that is common to both inputs (60 Hz noise and electromagnetic interference from other appliances in the room for example) will cancel out at the output. In reality, this value (called “common mode gain”) is never zero since the gains of the two amplifiers are not exactly equal. Its ability to reject signals that are common to both electrodes is stated by the ratio “gain/common mode gain” and it is called common mode rejection rate (CMRR). For surface EMG the CMRR should be in the area of 1,000,000:1 when a differential amplifier with a gain of 1,000 is used.
Using a differential amplifier is an excellent means of obtaining a signal that is free of interference. It is necessary though to process that signal even further in order to obtain a myographic signal that will be as pure as possible. This process is called filtering and the devices that are used are called filters. A high-pass filter will allow signals above a stated frequency to pass through. A low-pass filter on the other hand will allow signals bellow the stated frequency to pass through. The combination of a high-pass filter of 30 Hz and a low-pass filter of 500 Hz for example, will give a clean EMG signal between 30-500 Hz while all the other frequencies above and bellow will be selectively rejected. If a 60 Hz notch-filter is also used, the frequency of 60 Hz and therefor 60-cycle-noise will be specifically eliminated. Having gone through filtering, the signal still retains its A.C. bipolar characteristics and it needs to be converted into a D.C. pulsating signal. The device that is used to do that is call a “rectifier”.
A rectifier is a device that is able to keep the positive portion of the signal as it is, when at the same time it inverts the negative portion and gives it positive values. At its output then, it provides a signal that combines both the positive and the inverted-negative phases of the original signal. If, for example, the amplitude of the original signal ranges from -200 uV to +200 uV, when it goes through the rectifier it will range from 0 uV to +200 uV. In other words, the rectifier can be thought of as a device that is able to draw an axis of symmetry along the bipolar A.C. signal, fold it along that axis, and present it in a new, fast changing D.C. form where all the values range above zero. After the signal is rectified, it is fed into a device called an “integrator”.
The function of an integrator is to smooth the signal. It forms “envelopes” by plotting the line-of-best-fit among successive peaks of the signal, as per a given time interval. The time interval that determines the length of the envelops is called the time constant. Since it is connected in serial to the output of a rectifier, all of its output values will be non-zero and positive. So far, as the signal passes through a differential amplifier, a rectifier, and an integrator, it is shaped from a fast changing bipolar to a wave-shaped unipolar signal. In this final form it can fed into an A/D converter if it is to be interfaced with a microcomputer as part of the instrumentation, or it can fed directly to different audiovisual display configurations.
There are various types of audiovisual display devices used in biofeedback. Their evolution is characteristic of the galloping pace of progress in the field of electronics. Most of the commercial EMG biofeedback instrumentation has been developed for the average clinician that uses them for muscular relaxation within the parameters of psychotherapy or a stress management program. The displays of these devices are usually a small analog meter or a LED configuration in the front panel of the device. Some of the devices have been built with the presumption that the fancier their display, the more the patient will like them. This is all acceptable within the realm that a certain amount of suggestibility will facilitate the therapeutic process, but when it comes to neuromuscular reeducation the presumption is no longer valid. Three characteristics of the EMG biofeedback display are of extreme importance in neuromuscular reeducation: the intelligibility of the displayed signal, the preciseness of the signal, and the ability to display information on more than one muscle.
Intelligibility. The patient should be able to understand every bit of the feedback information that is displayed on the screen. Unnecessary information and fancy blinking lights distract the patient. Due to the nature of their disorder, some of these patients also have an organically based attention deficit. The display then, should linger within the limits of being simple enough to be absorbed, and stimulating enough to keep the patient focused.
Oscilloscopes should be given a special mention. An oscilloscope inspired the birth of EMG biofeedback when three investigators discovered that the display of the electromyogram facilitated the transmission of nerve impulses and as a consequence the motor units increased in frequency as well as in amplitude (Borsook, Billig, & Golseth, 1952). Undeniably, the sensitivity and precision that it provides is an asset in any laboratory. Its utility as a feedback device though, is compromised by the patient’s inability to assimilate the complexity of the wave.
Preciseness. The preciseness of the displayed signal becomes a function of two factors: the exactness of the signal and the delay of the instrumentation.
Signal exactness refers to the instrumentation’s ability to be able to display the detected signal as accurately as possible. If for example, the electrodes pickup a 1.0 uV value, the displayed value on the display should be as close to that as possible. Manufactures of biofeedback instruments refer to this characteristic as resolution or reliability. In some of the better commercially available devices this value averages at plus-or-minus 2.0 uV/sec. This simply means that if the electrodes picked up a 1.0 uV signal, on the display it could range anywhere between 0.0 – 3.0 uV/sec. Some of the best biofeedback configurations manage to keep their resolution within 0.5 uV/sec, and these are the best suited ones for neuromuscular reeducation. This range in large signals tends to be negligible, but in weak signals it can be detrimental.
The speed which the instrumentation processes the detected signal with, and presents it as feedback on the screen, is of utmost importance. The displayed signal will be processed by the brain in an active and immediate form. As the patient observes the feedback of his intended movement via the display, his/her brain determines whether he made a “hit” or a “miss”. It is important then, that he receives the feedback immediately, contingent upon the intended movement. Some of the fastest commercially available devices have a delay of as much as one third of a second per each channel. As yet, all these commercially available devices do their signal processing in series, which means that the more channels they can display feedback on, the slower their displaying speed.
A person who is learning how to play the piano can be used as an example to elucidate these two points. Let us suppose that he uses an electric piano, and that the sound that he produces by pressing the piano’s keys are fed to his ears via headphones. The headphones are connected serially to a delay device, and the delay of the sound can be manipulated. The longer the delay of the sound reaching his ears, the more difficult the conclusion of whether he hit the correct note or not a second or two ago. If in addition to the delay device some other device distorts the exactness of the sound that the depressed piano key should produce, his ability to learn piano will be thwarted even further. If for example, when he hits the “mi” note he hears any note between a “do” and a “sol”, it would be very similar to a patient producing a 3.0 uV signal but getting anything between 1.0 – 5.0 uV on the display. Steve Wolf (1978), one of the most successful EMG biofeedback clinicians, supports this idea firmly and calls for the therapist’s serious concern on the issue of preciseness. On… that note, the clinician should become aware of the jeopardies that are inherent in compromises of these two characteristics.
Multiple channels. Invariably, in all patients suitable for treatment with EMG biofeedback, feedback should be given not only from the muscle in question but from its antagonist as well. The reason is simple: no single muscle controls a joint by itself. There is always another muscle, the antagonist, which exerts corrective forces on the joint, and whose vectors stabilize the joint mechanically through various ranges of movement. This relationship does not change in pathological states: it is rather the reason for their occurrence. Hence, if paralysis in the arm of a stroke patient results from strong co-contractive patterns of the antagonistic muscles, correcting this maladaptive state becomes easier when feedback is provided from both muscles so that the patient can observe the state of events in both muscles.
Depending on the exact nature of the audiovisual displays, different types of “level detection” devises are implemented in the instrumentation. The purpose of these devices is to detect preset threshold levels and to provide continuous information of whether the threshold level has been met. By detecting electromechanically whether a certain response was a “hit” or a “miss”, they a) aid the patient’s judgment of the response, and b) reinforce the patient for making a response in the correct direction. In that sense alone, level detection devices serve as “cheerleaders” in terms of providing strong and immediate reinforcement, contingent upon the response.
One might ask then: Why is it important to implement such a device as part of a standard biofeedback instrumentation, since the patient himself can observe and decide whether the response was in the right direction or not? The answer to that is that the patient cannot always make the decision alone, at least not without investing a lot of mental energy to it. That mental energy could be employed into another area of concentration: recruitment of additional MUP’s. The process of recruitment will be discussed in the “clinical applications” section.
The Birth of EMG Biofeedback
The emergence of EMG biofeedback is not associated with a specific point in time. Electromyography itself did not emerge as a field until 1929, when Andrian and Bronk (1929a, 1929b) published in two landmark papers their first studies. These pioneer electromyographers showed that the electrical activity of individual muscles provided an accurate estimate of the actual function of the muscles. Incidentally, that was the same year that Sir Charles Sherrington and his group “discovered” the motor unit and began their extensive research on the neuromuscular system (Sherrington, (1929). Between that period and the 1950’s there have been only few reports on EMG (Gilson & Mills, 1941; Lindsley, 1935; Smith, 1934). The post-war period of the 1950’s was marked by very intense research efforts in electromyography, the focus of which was on the development of prosthetic limps that could be controlled electromechanically. Systematic investigations begun to explore the possibility of whether MUP’s could drive and control myoelectric prostheses. To ponder this likelihood, investigators had to understand how this control is exerted, and what is its exact nature. This research effort initiated work in two major areas: electromyographic kinesiology and single motor unit recruitment. Even though the former provides resourceful information in clinical EMG use, it is the later, through the exceptional work of John Basmajian, that shaped the future of EMG biofeedback.
The first publication announcing the observation that EMG feedback could facilitate recruitment of additional MUP’s on a paretic muscle, was that of Borsook, Billig, and Golheth (1952). Three years later, Marinacci (1955) published a book on his clinical observations and formulations regarding various neuromuscular diseases. In the same book he described his observations of the facilitating nature of EMG oscilloscopic information in recruiting MUP’s. In 1960, Marinacci and Horande published a series of case histories and discussed the competency of EMG feedback, as primary treatment, in the restoration of neuromuscular function in a variety of neurological conditions. Most authors identify this publication as the first one to appear in the “official” EMG biofeedback literature. Four years later, Andrews (1964) reports that 17 of his 20 chronically disabled patients were able to regain some elbow control after only five minutes of such treatment. Meanwhile, Basmajian (1963) with his newly developed fine-wire electrodes becomes the “demon of electromyography” and begins to publish a long list of articles on the conscious control of motor neurons, motor unit recruitment, and the electromyographic behavior of almost every muscle of the human body (Basmajian, 1957; 1959; 1963; 1965; 1967; 1972; 1974; Basmajian & Travill, 1961; Basmajian & Cross, 1971; Basmajian & White, 1973; Basmajian, Baenza, & Fabrigar, 1965).
While these events took place in the field of electromyography, some other interesting developments occurred in the field of psychology. In the 1920’s and 1930’s, Jacobson began developing his progressive muscle relaxation response on the basis of electromyographic data that he was collecting with his, rather primitive, EMG equipment (Jacobson 1929; 1934; 1938; 1939). Guided by the myoelectric responses that various body postures produced in his patients, he generated a protocol that facilitated systematic relaxation for a variety of psychoneurotic diseases. Other forms of systematic relaxation followed soon (Schultz, 1939; Schultz & Luthe, 1959), and a considerable amount of literature begun to accumulate on the uses of relaxation in the psychotherapeutic setting. At this point it was quite obvious that information about levels of muscular tension could facilitate decreases of the tension in a normal individual’s musculature, but it was debatable whether such information could be used more substantially in severe pathological states. Suggestions that feedback from visceral responses, or from muscles that were no more under the control of the central nervous system – i.e. in neuromuscular disorders- were strongly debated. Experts in the field of “the laws of learning” were extremely pessimistic (Schlosberg, 1937; Skinner, 1938; Mower, 1950). Meanwhile, Neal Miller, the person who was referred to as the “father of biofeedback” in the introductory section, was seeing things from a different perspective. Upon his return from his trip in the Soviet Union in the spring of 1960, he wrote:
It seems to me, however, entirely possible that this entire dichotomy is not the result of a basic difference in the fundamental properties of these two branches of the nervous system but, instead, is the result of the way in which the effectors of these systems are related to the environment under the normal conditions of life. Since the smooth muscles and glands usually do not have any instrumental effect on changing the external environment, Instrumental responses of this kind have not been previously reinforced. Therefore laboratory experiments on such learning not only require special instrumentation, but they also may start without the advantage of the enormous amount of transfer of training that usually is available to help the instrumental learning of those somatic responses usually selected for experimental study. I believe it is important to direct research toward this neglected problem in order to find out whether these two branches of the nervous system obey different laws” (Miller, 1961, p.835).
Miller means the differential response of the somatic vs. the autonomic nervous system in different conditions of instrumental learning. He refers specifically to the conditioning of visceral responses, which are, off course, under the control of the autonomic branch.
Miller’s words became action as he soon started developing the special instrumentation that was needed to make these “hidden” responses available to the external environment so that they could be reinforced and modified. In the years to follow he began communicating his findings from his Rockefeller University laboratory. (Miller & DiCara, 1967; 1968; DiCara & Miller, 1968a; 1968b; Miller, 1969 ). The work in Miller`s lab was the first to demonstrate that autonomic nervous system functions can be brought under voluntary control with the use of instrumental and operant conditioning procedures (DiCara & Miller, 1968).
The novelty and persuasiveness of his findings invited everyone to come aboard the bandwagon of the biofeedback panacea. At this writing, many laboratories in the country have investigated and evaluated the effectiveness of biofeedback in the treatment of a variety of disorders. As far as EMG biofeedback is concerned, as a rule of thumb, it is used by two separate groups of clinicians: those that deal with psychosomatic and/or psychogenic ailments, and those that are involved in the habilitation and/or rehabilitation of organically based disorders. EMG biofeedback is used by those in the first group for the treatment of disorders such as tension headaches, anxiety and relaxation training; by those in the second group for neuromuscular re-education in disorders such as stoke, spinal cord injuries, head trauma, as well as in the habilitation of writing disabilities, and developmental disabilities such as cerebral palsy.
There are two basic patterns of activity that characterize the clinical EMG picture of patients suffering from various neuromuscular disorders. Since this discussion will focus upon treatment utilizing surface electrodes, unless otherwise specified, all reported findings, observations, and measurements will refer to surface EMG measurements that are derived from superficial muscles, using surface silver/silver chloride electrodes. Both patterns are observed invariably along the spectrum of neuromuscular disorders, even though their underlying causes may vary.
The first pattern is that of decreased neuromotor activity at the muscle site. Rather than obtaining readings that range from 0-600 uV/sec, as in the normal muscle, the afflicted patient’s readings will range from 0 uV/sec to a fraction of the normal range and will depend on the extend of the lesion. Since for a muscle of normal bulk an average of 150-250 uV/sec is necessary to produce functional range of motion in antigravity movements, patients that fall below this value might not be able to produce voluntary movement. Readings that range around 0-50 uV/sec are quite common in some of these patients and contraction of the muscle might go undetected during a neurological or physical therapy evaluation that relies on palpation.
The second pattern is that of persistent co-contractions during attempts to produce movement. In some disorders, like a stroke for example, the co-contractions are rigid. This rigidity, which is due to the simultaneous firing of antagonistic muscle groups, leads to the characteristic paralysis that is produced by the mechanical opposition of muscle forces at the joint. When the opposing forces are not equal -either as a result of disproportional neurosignal loses or as a result of greater flaccidity and loss of strength of one muscle over the other- voluntary movement might be possible, but with limited range and considerable effort. The direction of the voluntary movement will be towards the vector of pull of the stronger muscle. In some other disorders, like a spinal cord injury for example, the co contractive pattern is manifested as spasticity. Spasticity is caused from excess activity at the spinal cord. Neural circuits that are responsible for mediation of recurrent and reciprocal inhibition become inactive due to the loss of descending input from higher centers. The gamma-loop comes directly under the control of spinal circuits that receive their input from the muscle afferents and grading of the muscle contractive forces is regulated by reflex mechanisms. The determinant factor of whether maladaptive co-contractive patterns will be manifested as rigidity or spasticity seems to be whether there is upper motor neuron or lower motor neuron involvement.
A third pattern of activity emerges as the two above mentioned patterns combine. Neurosignal losses are combined with persistent co-contractions, and voluntary movement production is impaired to analogous extends. This third, combined, pattern is the dominant picture that one encounters in clinical practice, since, almost always, the CNS injury does not focalize within a precise region with a specific function. The goal of biofeedback therapy then is to correct both of these problems within the shortest time period.
A muscle that is supplied by a limited degree of neurosignal innervation will become flaccid. Flaccidity becomes a relative function of the extent of the neurosignal innervation loss, the elapsed time since the loss occurred, and the degree to which the muscle is contracting over periods of time. In other words, a partially denervated muscle, over time, will atrophy to a degree that is proportional to the signal loss, but it will maintain its tone as long as some sort of stimulus triggers contractions of its fibers. Such mechanisms include co-contractive patterns that are elicited reflexively, tendinous reflexes, and electrical stimulation of the muscle. The treatment goal of EMG biofeedback, with such a partially denervated muscle, is to increase the amount of MUP’s that the patient can recruit voluntarily. This is accomplished by the patient as he/she recruits additional motor units by observing his/her amplified neuromotor responses on the display screen. As the patient attempts to increase activity in the muscle, the display reveals whether his/her attempt was successful or not. With weak signals, a “successful approximations towards the goal” approach is used, and the patient is reinforced individually for each small correct response towards the desired direction. This is where a level detection device becomes extremely valuable: the threshold detector of the device is more sensitive than the human eye in terms of distinguishing between a “hit” or a “miss”. The stimulus that is released by the level detector (sound or light) acts as a powerful reinforcer for the patient: it signals that the patient just made a hit, and it also reminds him/her that he is working in the right direction. Soon, the patient begins to be able to initiate firing of the additionally recruited motor neurons in the absence of feedback, and starts to integrate this newly acquired response in the production of functional range of motion. In addition, the newly augmented contractive activity (in the previously silent motor fibers) will lead to increases in the overall muscle tone and bulk.
Co-contractive patterns are disintegrated in a similar fashion. The threshold detector is set in a way that it can detect proportional differences in the activity of the two antagonists. For example, in a patient with foot-drop that is unable to dorsiflex due to a co-contractive pattern, the threshold detector is set to go off when the neurosignal activity of the tibialis anterior is just a hair greater than that of the gastrocnemius. As the patient’s ability to meet the threshold criterion improves, the criterion is set higher until the new level is met, and so forth. By successively reinforcing greater degrees of difference in the activity of the two muscles, the patient learns to fire the tibialis anterior actively while suppressing the activity of the gastrocnemius. The same approach is used with any other combination of co-contractive muscles.
In the third pattern of maladaptive activity, since it is the combination of the two above, the decision to commence treatment, towards the first or the second goal, is determined by the individual picture that each patient’s condition presents. It is always more desirable to proceed by recruiting additional motor units since the subsequent task of abolishing the co contractive pattern becomes easier. In some cases that the patient plateaus too soon though, the opposite direction may warrant better results.
Having reached the patient’s maximum potential of recruiting additional motor units in the muscle and having established a harmonious relationship between antagonistic muscle groups, the next step of the treatment sequence is that of putting the new neuromotor responses into functional use. As Meichenbaum notes, in clinical practice, biofeedback treatment is administered in the following stage sequence:
[span class=note]“…(a) training the client to develop an increased awareness of the specific physiological response, and (b) teaching the client to control his physiological response(s) voluntarily by means of biofeedback. The third and final stage involves the client`s employing his newly acquired voluntary controls in his natural environment” (Meichenbaum, 1976, p.202).[/span]
In the area of neuromuscular reeducation, EMG biofeedback has been applied successfully in stroke and head injury (Marinacci & Horande, 1960; Johnson & Garton, 1973; Amato, Hermsmeyer, & Kleinman, 1973; Swaan, van Wieringen, & Fokkema, 1974; Brudny et al., 1974; Basmajian, et al., 1975; Brudny et al., 1976; Basmajian et al., 1982); spinal cord injuries (Brudny et al, 1974; Seymour & Bassler, 1977; Nacht, Wolf, & Coogler, 1982; Brucker, 1983); cerebral palsy (Brudny, 1974; Wolpert & Wooldridge, 1975; Skrotzky, Galenstein, & Osternig, 1978; Asato, Twiggs, & Ellison, 1981); spasmodic torticollis (Cleeland, 1973; Brudny, Grynbaum, & Korein, 1974; Korein et al., 1976); nerve injury (Booker, Rubow, & Coleman, 1969; Kukulka, Brown, & Basmajian, 1975); poliomyelitis (Marinacci & Horande, 1960; Swann et al., 1974); Parkinson’s disease (Netsell & Cleeland, 1973; Nusselt & Legewie, 1975); writer’s cramp (Reavley, 1975); tendon transfers (Brudny et al., 1976); and Guillain-Barre syndrome (Cohen, Crouch, & Thompson, 1976; Ince, in press).
The EMG biofeedback literature has been reviewed and criticized in a number of books (Basmajian, 1977; 1978; Beatty & Legewie, 1977; Schwartz & Beatty, 1977; Birnbrauer & Kimmel, 1979; Olton & Noonberg, 1980; Yates, 1980; White & Tursky, 1982), and articles (Blanchard & Young, 1974; Inglis, Campbell, & Donald, 1976; Baker et al., 1977; Keefe & Surwit, 1978; Wolf, 1979; Basmajian, 1981; Kogeorgos & Scott, 1981). As it is frequently pointed out in these reviews, from a methodological point alone, the literature leaves a lot to be desired. Main flaws include lack of control or comparative groups, insufficient information reported by the investigators on a) treatment procedures, b) equipment used and electrode placement, and c) method of evaluation, number of sessions, and many other issues that originate from lack of consideration for the control of variables. In addition to the criticism that has been given so far, it should be added that many of these studies have put unrealistic demands on their patients. Many clinicians have attempted to treat their patients with primitive forms of instrumentation that, even though are more than sufficient for relaxation training and headache treatment, they are inadequate in neuromuscular reeducation. At the minimal level, the “preciseness” and “multiple channels” requirements have been seriously violated. In the course of examining the therapeutic effects of EMG biofeedback in neuromuscular reeducation, many investigators have speculated on the mechanisms of its action. Joseph Brudny and his group (1979) have theorized that a) “Information derived from processing and feedback from EMG during movement represents a substitute for disruptive kinesthetic transmission and processing related to brain insult.” and b) “… recovery may be mediated by substitute information reaching and being processed by brain areas unaffected by the insult.” (p. 818). In the light of one of their earlier communications (Brudny et al., 1977), (a) and (b) are conceptualized within the processes of override, where information from muscle receptors may be reaching the somatosensory cortex at a level higher than the locus of the injury and act directly upon corticospinal tract neurons; and by pass, where having reached the thalamus, the feedback information gets disbursed to structures of the subcortical level. Thus, by passing the somatosensory cortex, it flows from the brain stem motor nuclei directly on motorsensory feed-forward systems. In addition, influenced by the formulative basis of the ideas that have been discussed in this paper, this author speculates that, through the aid of the visual and auditory feedback, the patient as a system searches through a trial-and-error strategy for synapses that were rendered non-functional following the injury (since they lost their input or output primary connections). As EMG signal from these newly discovered synaptic connections summates on the display screen, the patient realizes that he/she is moving on the right direction and practices on ways that will ensure their use on future occasions (programming a new engram). As these previously unused or underused synapses gain functional status, depending on the amount of use that the system imposes on them for the production of motor movement, their axons proliferate, grow in size, and in turn reach for new synaptic connections within the web of the neuromuscular system.
Within their own fields, many other authors have reasoned along the same lines. For instance, Nichols (1982) has suggested the idea of collateral axonal sprouting to form new synapses, as well as the functional use of alternative synaptic pathways; Bach-y-Rita (1981) produced models of elimination of malfunctional active inhibitory activity and of unmasking of pre existing pathways in disorders of CNS lesions; and Geschwind (1974) contemplated the idea of transfer of function to intact neighboring neural structures. With the most recent experimental evidence at hand, it seems that the functional reorganization of neural circuitry follows two processes: an alteration in the efficacy of existing synaptic networks as a result of enduring changes in various synaptic events, and a morphological restructuring of the neurons and their connections.
Changes in synaptic events.
Learning can be described in simple terms as a behavioral adaptation that is related to previous experience. When the “learning theory of evolution” was advanced, the formation of the fight-or-flight response, for example, was proposed as the result of such learning. The question of whether the learning process involves some changes in the synaptic efficacy [a function of the excitatory postsynaptic potential (EPSP) amplitude and the manner that the postsynaptic neuron responds to it] has very often been implicated in attempts to examine the nature of the biobehavioral connection. Asking the same question, Byrne (1979) prepares us for an encounter with the unknown:
Indeed, a recurring problem with the cellular analysis of behavior has been to determine quantitatively the casual contributions which individual neurons and their biophysical properties make to the behavioral response. Ultimately, one would like to know to what degree a specific neural circuit and its biophysical properties, proposed for a given behavior, do actually account for that behavior. Stated in a slightly different way, to what degree can the whole of a behavior be fully described in terms of the sum of its neural components and their biophysical properties?” (p. 268).
In a discussion of “parts and wholes”, taken somewhat out of context, the philosopher Ernest Nagel writes:
The word “whole” sometimes refers to a property of an object or process, and “part” to some analogous property that stands to the first in certain specified relations…. Organic or “functional” have been defined as systems “the behavior of which is not determined by that of their individual elements, but where the part processes are themselves determined by the intrinsic nature of the whole.” (*1) What is distinctive of such systems, therefore, is that their parts do not act, and do not posses characteristics, independently of one another. On the contrary, their parts are supposed to be so related that any alteration in any one of them causes a change in all of the other parts.(*2) In consequence, functional wholes are also to be systems which cannot be built up out of elements by combining these later seriatim without producing changes in all those elements. Moreover, such wholes cannot have any part removed, without altering both that part and the remaining parts of the system.(*3) Accordingly, it is often claimed that a functional whole cannot be properly analyzed from an “additive point of view”; that is, the characteristic modes must be studied in situ, and the structure of activities of the whole cannot be inferred from properties displayed by its constituents in isolation from the whole. (1963, p. 137-147). (bold-face added by author).
(*1) Koffka, K. (1931). Gelstalt. In Encyclopedia for the Social Sciences, 6, 645. New York.
(*2) Lewin, K. (1937). Principles of Topological Psychology. New York: McGraw-Hill. (p. 218).
(*3)Koehler, W. (1924). Die Physischen Gestalten im Ruhe und im Sta-tionaeren Zustand. Braunschweig. (p. 42).
Probably, Nagel never realized the relevance of his writing in reference to mechanisms of learning and plasticity within the nervous system. Reinis and Goldman (1982) provide an exhaustive review of the experimental literature that is followed with an excellent discussion of this topic. According to these authors, some of the factors that are associated with synaptic plasticity and learning are: the process of neuronal hyperpolarization that can follow a train of impulses; synaptic efficacy of different neurons in different areas of the brain; the position of the neuronal surface (near the axon hillock: “detonator synapses”, on the tips of dendrites: “intensifying synapses”); the functional state of the brain (a term used to convey the existing conditions of excitation and inhibition of all brain neurons at any point in time); and the interaction of all or any of the above factors with each other. Of course, characteristic states that are associated with each of these factors depend on complex physico-chemical processes such as activity of pathways involving various neurotransmitters, functional changes in neurotransmitter production, and release and inactivation. In turn, alterations in each of these physico-chemical processes involve the participation of various biochemical mechanisms in both neurons and glial cells, which include respiration, energy metabolism, protein and lipid synthesis and degradation.
The morphological changes during recovery from injury are easier to quantify than the changes that occur in the synaptic events. Diamond et al. (1976) observed that when a sensory or motor fiber is injured, degeneration of its terminal portions and an immediate loss of function follow. Neighboring nerve fibers recognize the injury by sensing its byproducts within their chemical environment and they respond by spreading sprouts or branches from their axons. This mechanism that is called collateral sprouting, seems to result in functional compensation for the loses of the injured fiber. Within a sufficient time period, the axon of that injured nerve fiber regrows (Tsukahara, 1981) and advances towards the sites of its previous connections as if it were following the chemical messages of its own degeneration process. Under controlled laboratory conditions, (where there are usually quantitative and qualitative differences in the induced pathological conditions than those that are seen in clinical practice), axonal sprouting establishes functional synapses. Instead, in clinical pathology, a condition that is called synaptic repression is frequently encountered (Mark, 1980). Even though the synaptic connections are present they remain unused. It seems as if the brain has somehow erased them from its listing of functional synapses, and unaware of their presence it does not use them.
A different type of morphological change that neuronal plasticity is attributable to, is that of substitution. Most of recovery of function following CNS lesions is attributed to this process. Substitution relies on the redundancy of function that is to be found within different brain structures and at different levels. If, for example, a particular function is represented in the circuitry of more than one brain structure (due to a continuous evolutionary shift of the function towards higher levels), following an injury of the dominant controlling structure, subordinate structures that still retain cognizance of that function might take over (Davis, 1979).
Changes in synaptic density, terminal size and postsynaptic membrane thickening, volume of the neuropil (which is comprised of dendrites, blood capillaries, and glial cells), and associated increases in overall brain weight have been observed in learning situations where the environment was being manipulated (Rosenzweig & Bennett, 1978), and learning was evidenced in neural connections in the spinal cord (in conditions where the spinal cord was severed from the brain) of dogs (Shurrager & Culler, 1938) and cats (Patterson, 1975).
Aided by the use of EMG biofeedback displays, a patient that has suffered significant CNS damage and presents with a neuromuscular condition, is able to regain function. In conditions such as cerebral palsy, where the neuromuscular link has never been established at a functional level, patients can learn to produce and coordinate functional movement. What appears to be a task of extraordinary proportions, becomes a simple learning process due to the vast amount of neuronal plasticity that exists within the CNS, and which we are just starting to understand. The underlying mechanisms of these learning processes might even be much simpler from what we could ever expect. In this author’s laboratory, children as young as two and-a-half years of age, -some do not even speak english- with conditions such as head trauma, spina bifida, and cerebral palsy are learning how to gain function of muscles groups that they had never used before. They all work for a powerful reinforcer: M&M’s. Presented with a little piece of candy every time a tone is heard, they soon come to realize that in order to have another piece of candy come their way they must make that tone come on again. Indeed, very soon they master the task of producing the tone consistently. Then, the tone threshold is set in the direction of the desired response and a “successive approximation towards the goal” reinforcement procedure begins. The young child soon begins to recruit additional motor units and to inhibit antagonistic muscle groups. As the amount of newly recruited motor units reaches a critical level, contraction of the muscle as a whole begins to produce functional range of motion. Nothing is different in the procedures used so far (compared to treating an adult patient), other than the fact that a) there is no communication between the child and the therapist, and b) the child’s only motivation is getting another M&M. In other words, the child has no cognizance of the significance of “making the signal on the screen go higher”! Never the less, even in the absence of long term incentives, these neurosignal gains translate into function. Maybe then the neuromuscular system depends on a form of learning that is different than that of forming memories and cognitive skills. If this is indeed the case, there might be some other type of biofeedback equipment configuration that can be developed; one that would be more appropriate for young children.
In this situation, children are not interested in the visual display of their EMG since they do not understand it. They become interested in the auditory feedback, not because it appeals to them unconditionally but, because of its “magical” association with the colored little candies. Considering the facts from a child’s perspective, a toy that looks and feels like a pony and whose leverage with the horizontal plane, as well as its speed of rocking can be controlled by the muscle output of certain muscles of interest, might produce faster learning by being more reinforcing. Evident reinforcers are very important in working with children, since a child’s attention span depends on the degree of reinforcement that he/she receives from the environment.
Adults with various neuromuscular conditions work for different kinds of reinforcers: Regaining the function they once had. Unlike some children that never got old enough to realize what it means to have the function that they never had, adults are, usually, highly motivated towards their rehabilitation. In neuromuscular reeducation, EMG biofeedback compliments well other traditional modes of treatment. The National Institute of Mental Health report (Runck, 1980) noted:
[span class=note]“Without question the most widely accepted use of biofeedback is for the movement disorders. Usually done with electromyographic (EMG) feedback, the training is considered an adjunct to other procedures used in the rehabilitation of patients suffering with disabilities associated with neuromuscular disease” (p. 74)..[/span]
At the time of this writing, EMG biofeedback has been used only with groups of muscles that have an independent nerve supply. The work of Basmajian (1963, 1979) showed than humans can learn to fire, with an amazing degree of control, motor neurons within the same muscle. It could be possible then that patients can learn to contract independently different heads of the same muscle. If such learning is possible, some additional types of neuromuscular disorders could be treated with EMG biofeedback. One neuromuscular problem that falls in this category is a condition called patella misalignment or dislocating patella or subluxating patella. The three different names indicate a condition where the patella (kneecap) has a tendency to dislocate (pop out) from the knee joint. The correct alignment of the patella is coordinated by two of the four heads of the quadriceps muscle: the vastus medialis and the vastus lateralis. Due to a number of contributing pathogenetic factors, the vastus medialis atrophies and its pull on the patella weakens. Thus, the stronger vastus lateralis pulls the patella laterally and, if the right conditions exist, the patella dislocates. A subluxating patella is one that even though there is a strong pull from the vastus lateralis, the patella cannot ride over the lateral head of the femur.
A study then, could investigate whether the vastus medialis and the vastus lateralis can be trained individually in patients suffering from this painful condition. If such training is possible, there is a good probability of being able to treat such patients with EMG biofeedback.