Control Of Limb Prostheses Research Articles

Design of Embedded System for Multivariate Classification of Finger and Thumb Movements Using EEG Signals for Control of Upper Limb Prosthesis.

Brain Computer Interface (BCI) determines the intent of the user from a variety of electrophysiological signals. These signals, Slow Cortical Potentials, are recorded from scalp, and cortical neuronal activity is recorded by implanted electrodes. This paper is focused on design of an embedded system that is used to control the finger movements of an upper limb prosthesis using Electroencephalogram (EEG) signals. This is a follow-up of our previous research which explored the best method to classify three movements of fingers (thumb movement, index finger movement, and first movement). Two-stage logistic regression classifier exhibited the highest classification accuracy while Power Spectral Density (PSD) was used as a feature of the filtered signal. The EEG signal data set was recorded using a 14-channel electrode headset (a noninvasive BCI system) from right-handed, neurologically intact volunteers. Mu (commonly known as alpha waves) and Beta Rhythms (8–30 Hz) containing most of the movement data were retained through filtering using “Arduino Uno” microcontroller followed by 2-stage logistic regression to obtain a mean classification accuracy of 70%.

Myoelectric control of upper limb prostheses using linear discriminant analysis and multilayer perceptron neural network with back propagation algorithm

Myoelectric control of upper limb prostheses using linear discriminant analysis and multilayer perceptron neural network with back propagation algorithm

Myoelectric control of upper limb prostheses using linear discriminant analysis and multilayer perceptron neural network with back propagation algorithm

Electromyogram (EMG) signals or myoelectric signals (MESs) have two prominent areas in the field of biomedical instrumentation. EMG signals are primarily used to analyse the neuromuscular diseases such as myopathy and neuropathy. In addition, the EMG signal can be utilised in myoelectric control systems - where the external devices like upper limb prostheses, intelligent wheelchairs, and assistive robots can be controlled by acquiring surface EMG signals. The aim of present work is to obtain classification accuracy first by using linear discriminant analysis (LDA) classifier where principal component analysis (PCA) and uncorrelated linear discriminant analysis (ULDA) feature reduction techniques are used for upper limb prostheses control application. Next, the multilayer perceptron (MLP) neural network with back propagation algorithm is used to calculate the classification accuracy for upper limb prostheses control.

Continuous human locomotion identification for lower limb prosthesis control

Electromyogram (EMG) signal have immense possibility to identify the human activities and intention. This potential further extended towards prosthesis control using EMG signal from amputees residual limbs. The current study explores the pool of thirty-eight features containing time as well as frequency domain features. Also, the classification performance of six classifiers are compared to check the suitability for locomotion identification out of three daily life locomotions. An iterative greedy selection based feature selection algorithm is proposed. The results are further compared with pre-existing feature vectors for single muscle and multi muscle based locomotion identification approach. The results show that time domain features are best suited for locomotion identification, whereas SVM, LDA and NN classifiers emerged as suitable classifiers. The feature selection algorithm applied with multi muscle based locomotion identification approach shows better classification accuracy with lesser number of features. The obtained results are directly applicable for lower limb prosthesis control and man machine interface.

Factors associated with interest in novel interfaces for upper limb prosthesis control.

BackgroundSurgically invasive interfaces for upper limb prosthesis control may allow users to operate advanced, multi-articulated devices. Given the potential medical risks of these invasive interfaces, it is important to understand what factors influence an individual’s decision to try one.MethodsWe conducted an anonymous online survey of individuals with upper limb loss. A total of 232 participants provided personal information (such as age, amputation level, etc.) and rated how likely they would be to try noninvasive (myoelectric) and invasive (targeted muscle reinnervation, peripheral nerve interfaces, cortical interfaces) interfaces for prosthesis control. Bivariate relationships between interest in each interface and 16 personal descriptors were examined. Significant variables from the bivariate analyses were then entered into multiple logistic regression models to predict interest in each interface.ResultsWhile many of the bivariate relationships were significant, only a few variables remained significant in the regression models. The regression models showed that participants were more likely to be interested in all interfaces if they had unilateral limb loss (p ≤ 0.001, odds ratio ≥ 2.799). Participants were more likely to be interested in the three invasive interfaces if they were younger (p < 0.001, odds ratio ≤ 0.959) and had acquired limb loss (p ≤ 0.012, odds ratio ≥ 3.287). Participants who used a myoelectric device were more likely to be interested in myoelectric control than those who did not (p = 0.003, odds ratio = 24.958).ConclusionsNovel prosthesis control interfaces may be accepted most readily by individuals who are young, have unilateral limb loss, and/or have acquired limb loss However, this analysis did not include all possible factors that may have influenced participant’s opinions on the interfaces, so additional exploration is warranted.

Targeted Muscle Reinnervation for the Upper and Lower Extremity.

Myoelectric devices are controlled by electromyographic signals generated by contraction of residual muscles, which thus serve as biological amplifiers of neural control signals. Although nerves severed by amputation continue to carry motor control information intended for the missing limb, loss of muscle effectors due to amputation prevents access to this important control information. Targeted Muscle Reinnervation (TMR) was developed as a novel strategy to improve control of myoelectric upper limb prostheses. Severed motor nerves are surgically transferred to the motor points of denervated target muscles, which, after reinnervation, contract in response to neural control signals for the missing limb. TMR creates additional control sites, eliminating the need to switch the prosthesis between different control modes. In addition, contraction of target muscles, and operation of the prosthesis, occurs in reponse to attempts to move the missing limb, making control easier and more intuitive. TMR has been performed extensively in individuals with high-level upper limb amputations and has been shown to improve functional prosthesis control. The benefits of TMR are being studied in individuals with transradial amputations and lower limb amputations. TMR is also being investigated in an ongoing clinical trial as a method to prevent or treat painful amputation neuromas.

MEANS AND METHODS OF THE UPPER LIMBS PROSTHESES CONTROL

In this article means and methods of upper limb prostheses control are described. The analysis of merits and demerits of the invasive systems of registration of signals that control artificial limbs is carried out. The new technology of high-resolution electromyography (EMG HR) is described. The complexity of obtaining a signal without movement artifacts is considered as the main problem of non-invasive EMG HR. The possible ways of applying the EMG HR systems for rehabilitation, optimization of electrodes placement and reducing their quantity during the fitting process of upper limb prosthesis are proposed. The possibility of using dynamic forecasting of prosthesis movements in real time in EMG HR systems and improvement of intuitive control of the upper limb prostheses is highlighted.Ref. 23, fig. 8.

Abstract P25: A Caprine Model of a Novel Amputation Paradigm for Bi-directional Neural Control of a Bionic Limb

PURPOSE: No significant changes have been made to limb amputation surgical paradigms since the Civil War era. Traditional approaches are plagued by poor nerve treatment, which leads to neuromas and phantom pain, and complicates neural communication with advanced limb prostheses. In this study we present a caprine model of the Linked Residual Musculature Peripheral Nerve Interface (LRM-PNI), a surgical approach designed to reduce phantom pain and improve bi-directional neural control of a bionic limb. The key advancement in this architecture is the surgical coaptation of natively-innervated agonist-antagonist muscle pairs within the residuum. The benefits of this approach are two-fold. First, preservation of the mechanical coupling between agonist contraction and antagonist stretch allows physiologically-relevant muscle state feedback from a prosthesis through mechanical activation of native mechanoreceptors within these linked muscles. Second, by preserving functionality of these native mechanoreceptors, we limit the presence of non-meaningful phantom sensation, including phantom pain. We hypothesize that these advancements will allow for afferent proprioceptive sensation of a limb prosthesis through native neural pathways, which is crucial in both reflexive and volitional lower extremity control during gait. METHODS: A transtibial amputation was performed on the hindlimb of a goat, during which the lateral tarsal tunnel from the amputated ankle was grafted transversely to the medial surface of the tibia, serving as a “synovial pulley”. Distal ends of the lateral gastrocnemius and tibialis cranialis were then coapted to either end of this synovial pulley, allowing coupled motion of the agonist-antagonist muscle pair. The residual musculature was sensorized with implantable epimysial and intramuscular electrodes, as well as sonomicrometry crystals. Over a two-month period, electromyography and muscle-state measurements were recorded from the residuum while the animal ambulated without a prosthesis, and fluoroscopic evaluations of coupled motion in the presence of artificial muscle stimulation were carried out under anesthesia. The animal was then sacrificed, and the residuum was examined. RESULTS: Electromyography from residual musculature during gait indicate that the animal continued to send neural commands to the coapted agonist-antagonist muscle pair in the absence of a prosthetic limb device. Muscle-state measurements and fluoroscopic evaluations show clear coupled motion of the agonist-antagonist pair in the presence of both natural neural commands and artificial muscle stimulation. Integrated EMG from the agonist muscle was cross-correlated against muscle state recordings from the antagonist; it was determined that the average time delay from agonist muscle activation to antagonist muscle stretch was 93 ms, with a correlation coefficient of .73. Post-mortem examination of the residual tissues showed no signs of necrosis and no abundance of scarring. CONCLUSIONS: These findings indicate that the LRM-PNI architecture has the potential to provide coupled motion of agonist-antagonist muscle pairs, as in the pre-amputation physiological milieu. Guided by these findings, it is our expectation that further development of the LRM-PNI architecture will improve bi-directional neural control of advanced limb prostheses.

A murine model of a novel surgical architecture for proprioceptive muscle feedback and its potential application to control of advanced limb prostheses

Objective. Proprioceptive mechanisms play a critical role in both reflexive and volitional lower extremity control. Significant strides have been made in the development of bionic limbs that are capable of bi-directional communication with the peripheral nervous system, but none of these systems have been capable of providing physiologically-relevant muscle-based proprioceptive feedback through natural neural pathways. In this study, we present the agonist–antagonist myoneural interface (AMI), a surgical approach with the capacity to provide graded kinesthetic feedback from a prosthesis through mechanical activation of native mechanoreceptors within residual agonist–antagonist muscle pairs. Approach. (1) Sonomicrometery and electroneurography measurement systems were validated using a servo-based muscle tensioning system. (2) A heuristic controller was implemented to modulate functional electrical stimulation of an agonist muscle, using sonomicrometric measurements of stretch from a mechanically-coupled antagonist muscle as feedback. (3) One AMI was surgically constructed in the hindlimb of each rat. (4) The gastrocnemius-soleus complex of the rat was cycled through a series of ramp-and-hold stretches in two different muscle architectures: native (physiologically-intact) and AMI (modified). Integrated electroneurography from the tibial nerve was compared across the two architectures. Main results. Correlation between stretch and afferent signal demonstrated that the AMI is capable of provoking graded afferent signals in response to ramp-and-hold stretches, in a manner similar to the native muscle architecture. The response magnitude in the AMI was reduced when compared to the native architecture, likely due to lower stretch amplitudes. The closed-loop control system showed robustness at high stretch magnitudes, with some oscillation at low stretch magnitudes. Significance. These results indicate that the AMI has the potential to communicate meaningful kinesthetic feedback from a prosthetic limb by replicating the agonist–antagonist relationships that are fundamental to physiological proprioception.

Real-time gait event detection for lower limb amputees using a single wearable sensor.

This paper presents a rule-based real-time gait event/phase detection system (R-GEDS) using a shank mounted inertial measurement unit (IMU) for lower limb amputees during the level ground walking. Development of the algorithm is based on the shank angular velocity in the sagittal plane and linear acceleration signal in the shank longitudinal direction. System performance was evaluated with four control subjects (CS) and one transfemoral amputee (TFA) and the results were validated with four FlexiForce footswitches (FSW). The results showed a data latency for initial contact (IC) and toe off (TO) within a range of ± 40 ms for both CS and TFA. A delay of about 3.7 ± 62 ms for a foot-flat start (FFS) and an early detection of -9.4 ± 66 ms for heel-off (HO) was found for CS. Prosthetic side showed an early detection of -105 ± 95 ms for FFS whereas intact side showed a delay of 141 ±73 ms for HO. The difference in the kinematics of the TFA and CS is one of the potential reasons for high variations in the time difference. Overall, detection accuracy was 99.78% for all the events in both groups. Based on the validated results, the proposed system can be used to accurately detect the temporal gait events in real-time that leads to the detection of gait phase system and therefore, can be utilized in gait analysis applications and the control of lower limb prostheses.

Real-time evaluation of a myoelectric control method for high-level upper limb amputees based on homologous leg movements.

Electromyography-based gesture classification methods for control of advanced upper limb prostheses are limited either to individuals with amputations distal to the elbow or to those willing to undergo targeted muscle reinnervation surgery. Based on the natural similarity between gestures of the lower leg and the arm and on established methods in electromyography-based gesture classification, we propose a noninvasive system with which users control an upper limb prosthesis via homologous movements of the leg and foot. Eight inexperienced able-bodied subjects controlled a simulated robotic arm in a target achievement control (TAC) task with command of up to four degrees of freedom toward targets requiring one motion class. All subjects performed the task with analogous electromyography recording configurations on both the leg and the arm (as a benchmark), achieving slightly better performance with leg control overall. Only a brief demonstration of the arm-leg gesture mapping was necessary for subjects to perform the task, establishing the minimal training time required to begin using the control scheme. Our findings indicate that electromyography-based recognition of leg gestures may be a viable noninvasive prosthesis control option for high-level amputees.

A New Powered Lower Limb Prosthesis Control Framework Based on Adaptive Dynamic Programming

This brief presents a novel application of adaptive dynamic programming (ADP) for optimal adaptive control of powered lower limb prostheses, a type of wearable robots to assist the motor function of the limb amputees. Current control of these robotic devices typically relies on finite state impedance control (FS-IC), which lacks adaptability to the user's physical condition. As a result, joint impedance settings are often customized manually and heuristically in clinics, which greatly hinder the wide use of these advanced medical devices. This simulation study aimed at demonstrating the feasibility of ADP for automatic tuning of the twelve knee joint impedance parameters during a complete gait cycle to achieve balanced walking. Given that the accurate models of human walking dynamics are difficult to obtain, the model-free ADP control algorithms were considered. First, direct heuristic dynamic programming (dHDP) was applied to the control problem, and its performance was evaluated on OpenSim, an often-used dynamic walking simulator. For the comparison purposes, we selected another established ADP algorithm, the neural fitted Q with continuous action (NFQCA). In both cases, the ADP controllers learned to control the right knee joint and achieved balanced walking, but dHDP outperformed NFQCA in this application during a 200 gait cycle-based testing.

Spatial Representations in Local Field Potential Activity of Primate Anterior Intraparietal Cortex (AIP).

The execution of reach-to-grasp movements in order to interact with our environment is an important subset of the human movement repertoire. To coordinate such goal-directed movements, information about the relative spatial position of target and effector (in this case the hand) has to be continuously integrated and processed. Recently, we reported the existence of spatial representations in spiking-activity of the cortical fronto-parietal grasp network (Lehmann & Scherberger 2013), and in particular in the anterior intraparietal cortex (AIP). To further investigate the nature of these spatial representations, we explored in two rhesus monkeys (Macaca mulatta) how different frequency bands of the local field potential (LFP) in AIP are modulated by grip type, target position, and gaze position, during the planning and execution of reach-to-grasp movements. We systematically varied grasp type, spatial target, and gaze position and found that both spatial and grasp information were encoded in a variety of frequency bands (1–13Hz, 13–30Hz, 30–60Hz, and 60–100Hz, respectively). Whereas the representation of grasp type strongly increased towards and during movement execution, spatial information was represented throughout the task. Both spatial and grasp type representations could be readily decoded from all frequency bands. The fact that grasp type and spatial (reach) information was found not only in spiking activity, but also in various LFP frequency bands of AIP, might significantly contribute to the development of LFP-based neural interfaces for the control of upper limb prostheses.

A $24\,\mu \text{W}$, Batteryless, Crystal-free, Multinode Synchronized SoC “Bionode” for Wireless Prosthesis Control

We present a batteryless, crystal-free, time division multiple access (TDMA) synchronized multinode wireless body sensor node (WBSN) system-on-chip (SoC), referred to as a Bionode, for continuous and real-time telemetry of electromyograms (EMGs), enabling intuitive upper limb prosthesis control by an amputee. The SoC utilizes state of the art in supercapacitive RF energy harvesting, biosensing analog-front-end, switching-optimized SAR ADC, ultra-low-power RF transceiver, and clock circuits. The sensor node SoCs are time synchronized with a base station, mounted on the prosthetic arm, by using the ultra-low-power TDMA controller and receiver, and the digital core circuits. A 915 MHz broadcast RF signal is utilized to synthesize the carrier frequency of the transmitter. This along with the process and voltage compensated on-chip clock obviates the need for a bulky crystal oscillator, thus providing a low-cost and highly integrated solution to the WBSNs. The SoC is verified by capturing the EMG data from a healthy human body and consumes only $24\;{\upmu}\text{W}$ , while operating exclusively from the harvested RF energy. Implemented in a $0.18\;{\upmu}\text{m}$ CMOS process, the SoC occupies $2.025\;{\text {mm}}^{2}$ silicon area. The sensor node has an extremely low weight and physical dimensions, thanks to the flexible carbon nanotube (CNT) supercapacitor, electrically small antenna (ESA), and crystal-free operation of the SoC.

Stable force-myographic control of a prosthetic hand using incremental learning.

Force myography has been proposed as an appealing alternative to electromyography for control of upper limb prosthesis. A limitation of this technique is the non-stationary nature of the recorded force data. Force patterns vary under influence of various factors such as change in orientation and position of the prosthesis. We hereby propose an incremental learning method to overcome this limitation. We use an online sequential extreme learning machine where occasional updates allow continual adaptation to signal changes. The applicability and effectiveness of this approach is demonstrated for predicting the hand status from forearm muscle forces at various arm positions. The results show that incremental updates are indeed effective to maintain a stable level of performance, achieving an average classification accuracy of 98.75% for two subjects.

Exploiting arm posture synergies in activities of daily living to control the wrist rotation in upper limb prostheses: A feasibility study.

Although significant technological advances have been made in the last forty years, natural and effortless control of upper limb prostheses is still an open issue. Commercially available myoelectric prostheses present limited Degrees of Freedom (DoF) mainly because of the lack of available and reliable independent control signals from the human body. Thus, despite the crucial role that an actuated wrist could play in a transradial prosthesis in terms of avoiding compensatory movements, commercial hand prostheses present only manually adjustable passive wrists or actuated rotators controlled by (unnatural) sequential control strategies. In the present study we investigated the synergies between the humeral orientation with respect to the trunk and the forearm pronation/supination angles during the execution of a wide range of activities of daily living, in healthy subjects. Our results showed consistent postural synergies between the two selected body segments for almost the totality of the activities of daily living under investigation. This is a promising result because these postural synergies could be exploited to automatically control the wrist rotator unit in transradial prostheses improving the fluency and the dexterity of the amputee.

Task discrimination for non-weight-bearing movements using muscle synergies.

Myoelectric control of lower limb prostheses requires discrimination of task-specific muscle patterns. In this paper we present a method based on the notion of muscle synergies to discriminate between various non-weight-bearing movements such as knee extension/flexion, femur rotation in/out, tibia rotation in/out and ankle dorsiflexion/plantarflexion. Data is recorded from eight targeted muscle sites on the thigh. Non-negative matrix factorization is used to identify the muscle synergies using multiple features and estimation of electromyographic (EMG) patterns is done using non-negative least squares (NNLS). Classification accuracy for the movements involving the knee joint was higher than the movements involving the ankle joint. The proposed algorithm performs at par with the common machine learning algorithm Linear Discriminant Analysis (LDA) in offline analysis.

Context-Dependent Upper Limb Prosthesis Control for Natural and Robust Use.

Pattern recognition and regression methods applied to the surface EMG have been used for estimating the user intended motor tasks across multiple degrees of freedom (DOF), for prosthetic control. While these methods are effective in several conditions, they are still characterized by some shortcomings. In this study we propose a methodology that combines these two approaches for mutually alleviating their limitations. This resulted in a control method capable of context-dependent movement estimation that switched automatically between sequential (one DOF at a time) or simultaneous (multiple DOF) prosthesis control, based on an online estimation of signal dimensionality. The proposed method was evaluated in scenarios close to real-life situations, with the control of a physical prosthesis in applied tasks of varying difficulties. Test prostheses were individually manufactured for both able-bodied and transradial amputee subjects. With these prostheses, two amputees performed the Southampton Hand Assessment Procedure test with scores of 58 and 71 points. The five able-bodied individuals performed standardized tests, such as the box&block and clothes pin test, reducing the completion times by up to 30%, with respect to using a state-of-the-art pure sequential control algorithm. Apart from facilitating fast simultaneous movements, the proposed control scheme was also more intuitive to use, since human movements are predominated by simultaneous activations across joints. The proposed method thus represents a significant step towards intelligent, intuitive and natural control of upper limb prostheses.

Depth Sensing for Improved Control of Lower Limb Prostheses.

Powered lower limb prostheses have potential to improve the quality of life of individuals with amputations by enabling all daily activities. However, seamless ambulation mode recognition is necessary to achieve this goal and is not yet a clinical reality. Current intent recognition systems use mechanical and EMG sensors to estimate prosthesis and user status. We propose to complement these systems by integrating information about the environment obtained through the depth sensing. This paper presents the design, characterization, and the early validation of a novel stair segmentation system based on Microsoft Kinect. Static and dynamic tests were performed. A first experiment showed how the resolution of the depth camera affects the speed and the accuracy of segmentation. A second test proved the robustness of the algorithm to different staircases. Finally, we performed an online walking test with the stair segmentation and related measures recorded online at >5frames/s. Experimental results show that the proposed algorithm allows for an accurate estimate of distance, angle of intersection, number of steps, stair height, and stair depth for a set of stairs in the environment. The online test produced an estimate of whether the individual was approaching stairs in real time with approximately 98.8% accuracy.

Surveying the interest of individuals with upper limb loss in novel prosthetic control techniques

BackgroundNovel techniques for the control of upper limb prostheses may allow users to operate more complex prostheses than those that are currently available. Because many of these techniques are surgically invasive, it is important to understand whether individuals with upper limb loss would accept the associated risks in order to use a prosthesis.MethodsAn online survey of individuals with upper limb loss was conducted. Participants read descriptions of four prosthetic control techniques. One technique was noninvasive (myoelectric) and three were invasive (targeted muscle reinnervation, peripheral nerve interfaces, cortical interfaces). Participants rated how likely they were to try each technique if it offered each of six different functional features. They also rated their general interest in each of the six features. A two-way repeated measures analysis of variance with Greenhouse-Geisser corrections was used to examine the effect of the technique type and feature on participants’ interest in each technique.ResultsResponses from 104 individuals were analyzed. Many participants were interested in trying the techniques – 83 % responded positively toward myoelectric control, 63 % toward targeted muscle reinnervation, 68 % toward peripheral nerve interfaces, and 39 % toward cortical interfaces. Common concerns about myoelectric control were weight, cost, durability, and difficulty of use, while the most common concern about the invasive techniques was surgical risk. Participants expressed greatest interest in basic prosthesis features (e.g., opening and closing the hand slowly), as opposed to advanced features like fine motor control and touch sensation.ConclusionsThe results of these investigations may be used to inform the development of future prosthetic technologies that are appealing to individuals with upper limb loss.Electronic supplementary materialThe online version of this article (doi:10.1186/s12984-015-0044-2) contains supplementary material, which is available to authorized users.