Feedback control methods for task regulation by electricalstimulation of muscles

Three feedback control algorithms of varying complexity were compared for controlling three different tasks during electrical stimulation of muscles. Two controllers use stimulus pulse width (or recruitment) modulation to grade muscle force (the fixed parameter, first-order PW controller and the adaptive controller). The third controller varies both stimulus pulse width and period simultaneously for muscle force modulation (the PW/SP controller described in the comparison paper). The three tasks tested were isometric torque control, unloaded position tracking, and control of transitions between isometric and unloaded conditions. The first task involved the muscle recruitment nonlinearity. The second task added the effects of muscle length-tension and force-velocity nonlinearities. The third task included a sudden changes in external loading conditions. The comparative evaluation was carried out in an intact cat ankle joint with stimulation of tibialis anterior and medial gastrocnemius muscles. The simplest PW controller demonstrated robust control for all tasks. The PW/SP controller improved the performance of the PW controller significantly for control of isometric torque and load transition, but only slightly for control of unloaded joint position. However, the adaptive controller did not consistently achieve a significant improvement in performance compared with the PW controller for any task. Results suggest that muscle length-tension and force-velocity nonlinearities affect the performance of these controllers similarly within the tested ranges of movement amplitudes and speeds. Abrupt changes in the system, such as those due to recruitment nonlinearity and external loading transitions, tend to limit the performance of the adaptive controller. The study provides guidelines for choosing control algorithms for neural prostheses.     

Neuro-fuzzy extraction of angular information from muscle afferentsfor ankle control during standing in paraplegic subjects: an animalmodel

This paper is part of a project whose aim is the implementation of closed-loop control of ankle angular position during functional electrical stimulation (FES) assisted standing in paraplegic subjects using natural sensory information. In this paper, a neural fuzzy (NF) model is implemented to extract angular position information from the electroneurographic signals recorded from muscle afferents using cuff electrodes in an animal model. The NF model, named dynamic nonsingleton fuzzy logic system is a Mamdani-like fuzzy system, implemented in the framework of recurrent neural networks. The fuzzification procedure implemented was the nonsingleton technique which has been shown in previous works to be able to take into account the uncertainty in the data. The proposed algorithm was tested in different situations and was able to predict reasonably well the ankle angular trajectories especially for small excursions (as during standing) and when the stimulation sites are far from the registration sites. This suggests it may be possible to use activity from muscle afferents recorded with cuff electrodes for FES closed-loop control of ankle position during quite standing.     

Muscle recruitment with intrafascicular electrodes

We have studied muscle recruitment with Teflon-insulated, 25 microns diameter, Pt-Ir intrafasicular electrodes implanted in nerves innervating the gastrocnemius and soleus muscles of cats. The purpose of this study was to measure the performance of these bipolar electrodes, which had been designed to optimize their ability to record unit activity from peripheral nerves, as stimulating electrodes. Recruitment curves identified the optimal stimulus configuration as a biphasic rectangular pulse, with an interphase separation of about 500 microseconds and a duration of about 50 microseconds. The current required for a half-maximal twitch contraction was on the order of 50 microA. Current and charge densities needed for stimulation were well below levels believed to be safe for the tissue and electrode materials involved. When the spinal reflex pathway was interrupted by crushing the nerve, the force produced by a given stimulus changed in some cases, but not in others, implying that the spinal reflex contribution was not the same in all the implants. We conclude that intrafascicular recording electrodes are also a potentially valuable technology for functional neuromuscular stimulation, and warrant further development.     

The influence of stimulation pulse frequency on the generation ofjoint moments in the upper limb

Six muscles of the upper limb were stimulated transcutaneously. This communication reports the influence of the stimulation pulse frequency on the isometric joint moment generated by each muscle. A lower frequency of stimulation, the critical fusion frequency, was found for each muscle at which contractions ceased to be tetanic. This fusion frequency was correlated with muscle function. The magnitude of the joint moment was examined as a function of the stimulation pulse frequency for the six muscles tested. Parabolic curves were found to best fit this relationship; the magnitude of the moment reaching a maximum at typically 50 Hz, and often decreasing at higher frequencies. The slope of the linear portion of the relationship between the generated joint moment and the stimulation current intensity was shown to be a function of the stimulation pulse frequency. This function was found to be similar to the form of the joint moment versus pulse frequency curve. Fatigue curves were plotted at different stimulation frequencies; demonstrating reduced fatigue at lower stimulation frequencies. A model was presented of the fatigue curve as an exponential function of time. We conclude that a stimulation frequency of 15 Hz is optimal for the upper limb muscles with a working range of 15-50 Hz where stimulation frequency is one of the parameters used to modulate the muscle contraction force.     

Selective stimulation of peripheral nerve fibers using dualintrafascicular electrodes

The authors have studied activation of nerve fibers by pairs of Pt-Ir wire electrodes implanted within single fascicles of the nerve innervating the gastrocnemius muscle in cats. The purpose of this study was to determine if these intrafascicular electrodes can activate nerve fibers in different fascicles independently of each other and if they can also be used to activate separate subsets of axonal populations within a single fascicle. The average overlap of activated nerve fiber populations was 5.5% between fascicles and 27% within a fascicle, indicating that such selective activation is possible with these electrodes     

Sensory nerve recording for closed-loop control to restore motorfunctions

A method is developed for using neural recordings to control functional electrical stimulation (FES) to nerves and muscles. Experiments were done in chronic cats with a goal of designing a rule-based controller to generate rhythmic movements of the ankle joint during treadmill locomotion. Neural signals from the tibial and superficial peroneal nerves were recorded with cuff electrodes and processed simultaneously with muscular signals from ankle flexors and extensors in the cat's hind limb. Cuff electrodes are an effective method for long-term chronic recording in peripheral nerves without causing discomfort or damage to the nerve. For real-time operation the authors designed a low-noise amplifier with a blanking circuit to minimize stimulation artifacts. They used threshold detection to design a simple rule-based control and compared its output to the pattern determined using adaptive neural networks. Both the threshold detection and adaptive networks are robust enough to accommodate the variability in neural recordings. The adaptive logic network used for this study is effective in mapping transfer functions and therefore applicable for determination of gait invariants to be used for closed loop control in an FES system. Simple rule-bases will probably be chosen for initial applications to human patients. However, more complex FES applications require more complex rule-bases and better mapping of continuous neural recordings and muscular activity. Adaptive neural networks have promise for these more complex applications     

Analysis of passive elastic joint moments in paraplegics

In the functional electrical stimulation of the lower extremity of paraplegics to achieve standing and walking, a mathematical model describing the passive elastic joint moments is essential in order to implement model-based control algorithms. In a previous investigation of ten normal persons we had found significant coupling of passive, elastic joint moments between neighboring joints due to muscle groups that span both joints (biarticular muscles). Thus, we now investigated the biarticular coupling in six paraplegic patients. A comparison to the averaged results of the ten normal persons showed that while the biarticular joint moment coupling due to the gastrocnemius muscle was well preserved in all patients, the coupling due to the rectus femoris was greatly reduced and the coupling due to the hamstring muscle group was negligible. We offer pathophysiologically based explanations for these characteristic differences including the speculation that the predominantly extensor-type spasticity in our patients exercises mainly the anti-gravity muscles such as the gastrocnemius and the rectus femoris, while permitting greater atrophy of the hamstring muscle group. A previously presented double-exponential equation that predicts the joint moments under consideration of the neighboring joint angles could be fitted well to the experimental data.     

Classification of action potentials in multi-unit intrafascicularrecordings using neural network pattern-recognition techniques

Neural network pattern-recognition techniques were applied to the problem of identifying the sources of action potentials in multi-unit neural recordings made from intrafascicular electrodes implanted in cats. The network was a three-layer connectionist machine that used digitized action potentials as input. On average, the network was able to reliably separate 6 or 7 units per recording. As the number of units present in the recording increased beyond this limit, the number separable by the network remained roughly constant. The results demonstrate the utility of neural networks for classifying neural activity in multi-unit recordings     

Unsupported standing with minimized ankle muscle fatigue

In the past, limited unsupported standing has been restored in patients with thoracic spinal cord injury through open-loop functional electrical stimulation of paralyzed knee extensor muscles and the support of intact arm musculature. Here an optimal control system for paralyzed ankle muscles was designed that enables the subject to stand without hand support in a sagittal plane. The paraplegic subject was conceptualized as an underactuated double inverted pendulum structure with an active degree of freedom in the upper trunk and a passive degree of freedom in the paralyzed ankle joints. Control system design is based on the minimization of a cost function that estimates the effort of ankle joint muscles via observation of the ground reaction force position, relative to ankle joint axis. Furthermore, such a control system integrates voluntary upper trunk activity and artificial control of ankle joint muscles, resulting in a robust standing posture. Figures are shown for the initial simulation study, followed by disturbance tests on an intact volunteer and several laboratory trials with a paraplegic person. Benefits of the presented methodology are prolonged standing sessions and in the fact that the subject is able to maintain voluntary control over upper body orientation in space, enabling simple functional standing.     

A method to effect physiological recruitment order in electricallyactivated muscle

A new stimulation method has been utilized to achieve physiological recruitment order of small-to-large motor units in electrically activated muscles. The use of quasitrapezoidal-shaped pulses and a tripolar cuff electrode made selective activation of small motor axons possible, thus recruiting slow-twitch, fatigue-resistant muscle units before fast-twitch, fatigable units in a heterogeneous muscle. Isometric contraction force from the medial gastrocnemius muscle was measured in five cats. The physiological recruitment order was evidenced by larger twitch widths at lower force levels and smaller twitch widths at higher force levels in the muscles tested. In addition, force modulation process was more gradual and fused contractions were obtained at lower stimulation frequencies when the new stimulation method was employed. Furthermore, muscles activated by the new method were more fatigue-resistant under repetitive activation at low force levels. This stimulation method is simpler to implement and has fewer adverse effects on the neuromuscular system than previous blocking methods. Therefore, it may have applications in future functional neuromuscular stimulation systems.     

Maximum Likelihood Identification of a Posture Control System

To identify characteristics of a sensory feedback path of a posture control system without external disturbances, the maximum likelihood identification method was applied to data of an antero-posterior sway during quiet stance. A subject stood on a force plate and the sway angle of a body (input) was measured by a position sensor camera and the ankle joint moment (output) was measured by the force plate. Using data of 500 input-output pairs sampled every 0.1 s, parameters of the sensory feedback path were estimated and frequency response functions were calculated. These showed derivative characteristics that are necessary for stabilization of the posture control system. Gain characteristics under the closed eyes condition tended to be greater than those under the open eyes condition.     

Action potential classification with dual channel intrafascicularelectrodes

Using recordings of peripheral nerve activity made with carbon fiber intrafascicular electrodes, the authors compared the performance of three different recording techniques (single channel, differential, and dual channel) and four different unit classification methods (linear discriminant analysis, template matching, a novel time amplitude windowing technique, and neural networks) in terms of errors in waveform classification and artifact rejection. Dual channel recording provided uniformly superior unit separability, neural networks gave the lowest classification error rates, and template matching had the best artifact rejection performance     

Elbow extension in the C5 quadriplegic using functionalneuromuscular stimulation

A system has been designed to provide overhead reach in C5/6 quadriplegic subjects using functional neuromuscular stimulation (FNS) for control of the triceps muscle. The system uses the position of the arm in space as the input command, relieving the user from having to supply a conscious command signal. By measuring the position of the arm, the magnitude of the gravitational and passive torques opposing elbow extension can be calculated. This torque is counteracted by electrical activation of the triceps muscle, with the appropriate stimulus parameters determined from the recruitment characteristics of each electrode. Sufficient stimulus is applied to produce full elbow extension. Intermediate elbow angles are achieved using voluntary elbow flexor torque to counteract the effects of the stimulation. System performance was tested in two subjects. Subjects were asked to reach targets with and without stimulation, with loads up to 500 g in the hand. Using the FNS system, subjects were able to successfully reach the target positions above the horizontal that were inaccessible without stimulation.     

An interactive graphics-based model of the lower extremity to studyorthopaedic surgical procedures

We have developed a model of the human lower extremity to study how surgical changes in musculoskeletal geometry and musculotendon parameters affect muscle force and its moment about the joints. The lines of action of 43 musculotendon actuators were defined based on their anatomical relationships to three-dimensional bone surface representations. A model for each actuator was formulated to compute its isometric force-length relation. The kinematics of the lower extremity were defined by modeling the hip, knee, ankle, subtalar, and metatarsophalangeal joints. Thus, the force and joint moment that each musculotendon actuator develops can be computed for any body position. The joint moments calculated with the model compare well with experimentally measured isometric joint moments. We developed a graphical interface to the model that allows the user to visualize the musculoskeletal geometry and to manipulate the model parameters to study the biomechanical consequences of orthopaedic surgical procedures. For example, tendon transfer and lengthening procedures can be simulated by adjusting the model parameters according to various surgical techniques. Results of the simulated surgeries can be analyzed quickly in terms of postsurgery muscle forces and other biomechanical variables. Just as interactive graphics have enhanced engineering design and analysis, we have found that graphics-based musculoskeletal models are effective tools for designing and analyzing surgical procedures.     

Control of end-point forces of a multijoint limb by functionalneuromuscular stimulation

A multivariable feedback controller was designed and tested for regulating the magnitude and orientation of the force vector at the end point of a multijoint limb in contact with an isometric load. The force vector was produced by electrical stimulation of muscles. To achieve arbitrary control of end-point force magnitude and orientation, two coupling issues must be dealt with by the control system. First, there is a geometric coupling between the end-point force vector and joint torques. The amplitude and orientation of the force vector depend on the limb geometry. Second, torques at two joints may be coupled due to activation of muscles that cross them (biarticular coupling). To eliminate the geometric coupling, a transformation of controller error from the Cartesian space to the joint space was employed. A multivariable proportional-plus-integral (PI) control law was used to calculate muscle activation based on the transformed controller error. Centralized and decentralized controls were investigated for decoupling the effects of biarticular muscles. The results obtained from cat experiments showed that the magnitude and orientation of the end-point forces of the cat hindlimb could be regulated by this controller. In the presence of strong biarticular coupling, centralized control yielded better performance than decentralized control during transient responses. Both control strategies could decouple the biarticular muscle at steady state. When no biarticular coupling was present, centralized control sometimes performed worse than decentralized control. This is the first step in the simultaneous control of multiple joints by functional neuromuscular stimulation (FNS). The controller has broad potential applications in FNS neural prostheses.     

Closed-Loop Positioning of Hemiplegic Patient''s Joint by Means of Functional Electrical Stimulation

We report about the results on positioning of the ankle joint of a hemiplegic patient and of normal experimental subjects, using functional electrical stimulation (FES) of antagonistic muscle groups and position feedback. Such a controller is intended to be used as the execution level of a multilevel orthotic system. The synthesis of the controller has been made with the aid of the components of an analog hybrid computer. Experimental results show that the static and dynamic properties of regulated movements are not essentially inferior to the properties of voluntary movements. There are stated technological problems that will have to be solved before the controller can be clinically applied for rehabilitation purposes. At the present state of the art, the poor technology of FES, the unsolved problem of fatigue with FES muscles, size, price, and bad cosmetic effect prevent the position controller from being clinically applicable.     

Ambulatory assessment of ankle and foot dynamics

Ground reaction force (GRF) measurement is important in the analysis of human body movements. The main drawback of the existing measurement systems is the restriction to a laboratory environment. This paper proposes an ambulatory system for assessing the dynamics of ankle and foot, which integrates the measurement of the GRF with the measurement of human body movement. The GRF and the center of pressure (CoP) are measured using two six-degrees-of-freedom force sensors mounted beneath the shoe. The movement of foot and lower leg is measured using three miniature inertial sensors, two rigidly attached to the shoe and one on the lower leg. The proposed system is validated using a force plate and an optical position measurement system as a reference. The results show good correspondence between both measurement systems, except for the ankle power estimation. The root mean square (RMS) difference of the magnitude of the GRF over 10 evaluated trials was (0.012 +/- 0.001) N/N (mean +/- standard deviation), being (1.1 +/- 0.1)% of the maximal GRF magnitude. It should be noted that the forces, moments, and powers are normalized with respect to body weight. The CoP estimation using both methods shows good correspondence, as indicated by the RMS difference of (5.1 +/- 0.7) mm, corresponding to (1.7 +/- 0.3)% of the length of the shoe. The RMS difference between the magnitudes of the heel position estimates was calculated as (18 +/- 6) mm, being (1.4 +/- 0.5)% of the maximal magnitude. The ankle moment RMS difference was (0.004 +/- 0.001) Nm/N, being (2.3 +/- 0.5)% of the maximal magnitude. Finally, the RMS difference of the estimated power at the ankle was (0.02 +/- 0.005) W/N, being (14 +/- 5)% of the maximal power. This power difference is caused by an inaccurate estimation of the angular velocities using the optical reference measurement system, which is due to considering the foot as a single segment. The ambulatory system considers separate heel and forefoot segments, thus allowing an additional foot moment and power to be estimated. Based on the results of this research, it is concluded that the combination of the instrumented shoe and inertial sensing is a promising tool for the assessment of the dynamics of foot and ankle in an ambulatory setting.     

A modeling study of nerve fascicle stimulation

A nerve stimulation model has been developed, incorporating realistic cross-sectional nerve geometries and conductivities. The potential field in the volume conductor was calculated numerically using the variational method. Nerve fiber excitation was described by the model of McNeal. Cross-sectional geometries of small monofascicular rat common peroneal nerve and multifascicular human deep peroneal nerve were taken as sample geometries. Selective stimulation of a fascicle was theoretically analyzed for several electrode positions: outside the nerve, in the connective tissue of the nerve, and inside a fascicle. The model results predict that the use of intraneural or even intrafascicular electrodes is necessary for selective stimulation of fascicles not lying at the surface of the nerve. Model predictions corresponded with experimental results of Veltink et al. on intrafascicular and extraneural stimulation of rat common peroneal nerve and to results of McNeal and Bowman on muscle selective stimulation in multifascicular dog sciatic nerve using an extraneural multielectrode configuration.     

Instrumentation for ENG and EMG recordings in FES systems

An electronic circuit for analog processing of neural (electroneurogram or ENG) and muscular (electromyogram or EMG) signals in functional electrical stimulation (FES) systems is described. The basic circuit consists of a low-noise gated preamplifier, bandpass filter, amplifier, and a blanking circuit to minimize stimulation artifacts during electrical stimulation. This device was tested in chronic recordings using a triphasic cliff electrode for nerves and epimysial electrodes for muscles in the hind limbs of cats. The device was used for nerve recordings in the presence of electrical stimulation of muscles in the same leg. The recordings showed rejection of stimulation and muscle (M-wave) artifacts, while retaining the information of interest     

Information contained in sensory nerve recordings made withintrafascicular electrodes

Multiunit recordings were made in anesthetized cats with chronically implanted intrafascicular electrodes over a period of six months. Neural signals recorded with these electrodes consisted of activity in sensory fibers innervating a variety of cutaneous mechanoreceptors. Mechanical stimuli were used to selectively activate individual nerve fibers, and the receptive field and receptor type were identified for each unit. Over a period of six months, there was a net shift in the recorded population, but the electrodes continued to provide a representative sample of the activity in the fascicle as a whole. The total number of units from which activity could be recorded remained roughly constant with time, and individual units persisted in the recordings for up to six months. These results indicate that intrafascicular electrodes could be used to sample information carried by individual somatosensory fibers on a long term basis.