Prosthesis Control Research Articles

2자유도 저관성 능동 발목 의족 개발 및 비평탄면 보행을 위한 임피던스 제어

In this paper, a lightweight low-inertia 2-DOF ankle prosthesis is presented. This prosthesis is designed to reduce the heterogeneity of amputees and enable them to walk comfortably with low effort. The compact size and low height of this prosthesis means that it can be widely used by amputees with various physical conditions. The proposed ankle prosthesis has two active joints in the sagittal and the coronal planes, and these joints are actuated by a parallel linkage mechanism, which helps amputees to walk normally and maintain balance when stepping on obstacles or uneven ground. A unique spring mechanism helps with propulsion without increasing the distal mass. This paper describes the kinematics of the proposed prosthesis and joint-level impedance control and presents a balancing control algorithm based on the human gait cycle. The height and weight of the developed ankle prosthesis are 220 ㎜ and 2.25 ㎏, respectively, which are comparable to those of commercial 1-DOF prostheses. The experiments conducted using the proposed prosthesis on even and uneven terrains demonstrated that the proposed 2-DOF prosthesis appropriately adapts to various ground conditions and consistently assists users with balance and propulsion.

Toward higher-performance bionic limbs for wider clinical use.

Most prosthetic limbs can autonomously move with dexterity, yet they are not perceived by the user as belonging to their own body. Robotic limbs can convey information about the environment with higher precision than biological limbs, but their actual performance is substantially limited by current technologies for the interfacing of the robotic devices with the body and for transferring motor and sensory information bidirectionally between the prosthesis and the user. In this Perspective, we argue that direct skeletal attachment of bionic devices via osseointegration, the amplification of neural signals by targeted muscle innervation, improved prosthesis control via implanted muscle sensors and advanced algorithms, and the provision of sensory feedback by means of electrodes implanted in peripheral nerves, should all be leveraged towards the creation of a new generation of high-performance bionic limbs. These technologies have been clinically tested in humans, and alongside mechanical redesigns and adequate rehabilitation training should facilitate the wider clinical use of bionic limbs.

Intuitive real-time control strategy for high-density myoelectric hand prosthesis using deep and transfer learning

Myoelectric hand prostheses offer a way for upper-limb amputees to recover gesture and prehensile abilities to ease rehabilitation and daily life activities. However, studies with prosthesis users found that a lack of intuitiveness and ease-of-use in the human-machine control interface are among the main driving factors in the low user acceptance of these devices. This paper proposes a highly intuitive, responsive and reliable real-time myoelectric hand prosthesis control strategy with an emphasis on the demonstration and report of real-time evaluation metrics. The presented solution leverages surface high-density electromyography (HD-EMG) and a convolutional neural network (CNN) to adapt itself to each unique user and his/her specific voluntary muscle contraction patterns. Furthermore, a transfer learning approach is presented to drastically reduce the training time and allow for easy installation and calibration processes. The CNN-based gesture recognition system was evaluated in real-time with a group of 12 able-bodied users. A real-time test for 6 classes/grip modes resulted in mean and median positive predictive values (PPV) of 93.43% and 100%, respectively. Each gesture state is instantly accessible from any other state, with no mode switching required for increased responsiveness and natural seamless control. The system is able to output a correct prediction within less than 116 ms latency. 100% PPV has been attained in many trials and is realistically achievable consistently with user practice and/or employing a thresholded majority vote inference. Using transfer learning, these results are achievable after a sensor installation, data recording and network training/fine-tuning routine taking less than 10 min to complete, a reduction of 89.4% in the setup time of the traditional, non-transfer learning approach.

A compact system for simultaneous stimulation and recording for closed-loop myoelectric control

BackgroundDespite important advancements in control and mechatronics of myoelectric prostheses, the communication between the user and his/her bionic limb is still unidirectional, as these systems do not provide somatosensory feedback. Electrotactile stimulation is an attractive technology to close the control loop since it allows flexible modulation of multiple parameters and compact interface design via multi-pad electrodes. However, the stimulation interferes with the recording of myoelectric signals and this can be detrimental to control.MethodsWe present a novel compact solution for simultaneous recording and stimulation through dynamic blanking of stimulation artefacts. To test the system, a feedback coding scheme communicating wrist rotation and hand aperture was developed specifically to stress the myoelectric control while still providing meaningful information to the subjects. Ten subjects participated in an experiment, where the quality of closed-loop myoelectric control was assessed by controlling a cursor in a two degrees of freedom target-reaching task. The benchmark performance with visual feedback was compared to that achieved by combining visual feedback and electrotactile stimulation as well as by using electrotactile feedback only.ResultsThere was no significant difference in performance between visual and combined feedback condition with regards to successfully reached targets, time to reach a target, path efficiency and the number of overshoots. Therefore, the quality of myoelectric control was preserved in spite of the stimulation. As expected, the tactile condition was significantly poorer in completion rate (100/4% and 78/25% for combined and tactile condition, respectively) and time to reach a target (9/2 s and 13/4 s for combined and tactile condition, respectively). However, the performance in the tactile condition was still good, with no significant difference in path efficiency (38/8%) and the number of overshoots (0.5/0.4 overshoots), indicating that the stimulation was meaningful for the subjects and useful for closed-loop control.ConclusionsOverall, the results demonstrated that the developed system can provide robust closed-loop control using electrotactile stimulation. The system supports different encoding schemes and allows placing the recording and stimulation electrodes next to each other. This is an important step towards an integrated solution where the developed unit will be embedded into a prosthetic socket.

Recent Advances in Myoelectric Control for Finger Prostheses for Multiple Finger Loss

The loss of one or multiple fingers can lead to psychological problems as well as functional impairment. Various options exist for replacement and restoration after hand or finger loss. Prosthetic hand or finger prostheses improve esthetic outcomes and the quality of life for patients. Myoelectrically controlled hand prostheses have been used to attempt to produce different movements. The available articles (original research articles and review articles) on myoelectrically controlled finger/hand prostheses from January 1922 to February 2021 in English were reviewed using MEDLINE/PubMed, Web of Science, and ScienceDirect resources. The articles were searched using the keywords “finger/hand loss”, “finger prosthesis”, “myoelectric control”, and “prostheses” and relevant articles were selected. Myoelectric or electromyography (EMG) signals are read by myoelectrodes and the signals are amplified, from which the muscle’s naturally generated electricity can be measured. The control of the myoelectric (prosthetic) hands or fingers is important for artificial hand or finger movement; however, the precise control of prosthetic hands or fingers remains a problem. Rehabilitation after multiple finger loss is challenging. Implants in finger prostheses after multiple finger loss offer better finger prosthesis retention. This article presents an overview of myoelectric control regarding finger prosthesis for patients with finger implants following multiple finger loss.

Emerging Wearable Interfaces and Algorithms for Hand Gesture Recognition: A Survey.

Hands are vital in a wide range of fundamental daily activities, and neurological diseases that impede hand function can significantly affect quality of life. Wearable hand gesture interfaces hold promise to restore and assist hand function and to enhance human-human and human-computer communication. The purpose of this review is to synthesize current novel sensing interfaces and algorithms for hand gesture recognition, and the scope of applications covers rehabilitation, prosthesis control, exoskeletons for augmentation, sign language recognition, human-computer interaction, and user authentication. Results showed that electrical, mechanical, acoustical/vibratory, and optical sensing were the primary input modalities in gesture recognition interfaces. Two categories of algorithms were identified: 1) classification algorithms for predefined, fixed hand poses and 2) regression algorithms for continuous finger and wrist joint angles. Conventional machine learning algorithms, including linear discriminant analysis, support vector machines, random forests, and non-negative matrix factorization, have been widely used for a variety of gesture recognition applications, and deep learning algorithms have more recently been applied to further facilitate the complex relationship between sensor signals and multi-articulated hand postures. Future research should focus on increasing recognition accuracy with larger hand gesture datasets, improving reliability and robustness for daily use outside of the laboratory, and developing softer, less obtrusive interfaces.

A new approach of inducing proprioceptive illusion by transcutaneous electrical stimulation

BackgroundNeurotraumas or neurodegenerative diseases often result in proprioceptive deficits, which makes it challenging for the nervous system to adapt to the compromised sensorimotor conditions. Also, in human machine interactions, such as prosthesis control and teleoperation, proprioceptive mismatch limits accuracy and intuitiveness of controlling active joints in robotic agents. To address these proprioceptive deficits, several invasive and non-invasive approaches like vibration, electrical nerve stimulation, and skin stretch have been introduced. However, proprioceptive modulation is still challenging as the current solutions have limitations in terms of effectiveness, usability, and consistency. In this paper, we propose a new way of modulating proprioception using transcutaneous electrical stimulation. We hypothesized that transcutaneous electrical stimulation on elbow flexor muscles will induce illusion of elbow joint extension.MethodEight healthy human subjects participated in the study to test the hypothesis. Transcutaneous electrodes were placed on different locations targeting elbow flexor muscles on human subjects and experiments were conducted to identify the best locations for electrode placement, and best electrical stimulation parameters, to maximize induced proprioceptive effect. Arm matching experiments and Pinocchio illusion test were performed for quantitative and qualitative analysis of the observed effects. One-way repeated ANOVA test was performed on the data collected in arm matching experiment for statistical analysis.ResultsWe identified the best location for transcutaneous electrodes to induce the proprioceptive illusion, as one electrode on the muscle belly of biceps brachii short head and the other on the distal myotendinous junction of brachioradialis. The results for arm-matching and Pinocchio illusion tests showed that transcutaneous electrical stimulation using identified electrode location and electrical stimulation parameters evoked the illusion of elbow joint extension for all eight subjects, which supports our hypothesis. On average, subjects reported 6.81° angular illusion of elbow joint extension in arm-matching tests and nose elongated to 1.78 × height in Pinocchio illusion test.ConclusionsTranscutaneous electrical stimulation, applied between the the synergistic elbow flexor muscles, consistently modulated elbow joint proprioception with the illusion of elbow joint extension, which has immense potential to be translated into various real-world applications, including neuroprosthesis, rehabilitation, teleoperation, mixed reality, and etc.

Implementation of Scale Invariant Feature Transform detector on FPGA for low‐power wearable devices for prostheses control

SummaryIn this paper we describe an FPGA implementation of the Scale Invariant Feature Transform (SIFT) algorithm. The FPGA is required as its a lightweight device which makes it ideal for vision‐guided hybrid neuro‐prostheses utilised for upper limbs replacement. SIFT point detection is needed for computation of coordinates of the object‐to‐grasp in a wearable multi‐camera system. A modified SIFT algorithm is proposed and an implementation of it into C/C++ language on Xilinx ZCU102 FPGA board. The proposed hybrid hardware/software solution is compared to other hardware or hybrid implementations of the SIFT algorithm and with the baseline software detector OpenSIFT. The algorithm optimised for FPGA gives an average precision of 0.84 and the average recall of 0.94 in SIFT‐point detection compared to the baseline.The proposed solution has lower dissipated power than other solutions like CPU or GPU, and it has better computational speed.This solution allows for processing medium‐sized images in real‐time with low power consumption.

Shortcomings of human-in-the-loop optimization of an ankle-foot prosthesis emulator: a case series

Human-in-the-loop optimization allows for individualized device control based on measured human performance. This technique has been used to produce large reductions in energy expenditure during walking with exoskeletons but has not yet been applied to prosthetic devices. In this series of case studies, we applied human-in-the-loop optimization to the control of an active ankle-foot prosthesis used by participants with unilateral transtibial amputation. We optimized the parameters of five control architectures that captured aspects of successful exoskeletons and commercial prostheses, but none resulted in significantly lower metabolic rate than generic control. In one control architecture, we increased the exposure time per condition by a factor of five, but the optimized controller still resulted in higher metabolic rate. Finally, we optimized for self-reported comfort instead of metabolic rate, but the resulting controller was not preferred. There are several reasons why human-in-the-loop optimization may have failed for people with amputation. Control architecture is an unlikely cause given the variety of controllers tested. The lack of effect likely relates to changes in motor adaptation, learning, or objectives in people with amputation. Future work should investigate these potential causes to determine whether human-in-the-loop optimization for prostheses could be successful.

Neuromechanical Signal-Based Parallel and Scalable Model for Lower Limb Movement Recognition

Individual who have lost their lower limb because of amputation can use the prosthesis to restore daily living activities. The amputee intent recognition during locomotion modes can be used as source to control lower limb prosthesis. Due to continuous data recording from multiple sensors, the timely recognition of activities of daily living have become a challenging issue for traditional technology and conventional machine learning algorithms. This work hypothesize that parallel discriminant features can be learned from large amount of data generated by aggregating the neuromechanical signals from multiple subjects with parallel and distributed computing platform. Consequently, this paper apply three classifiers including support vector machine, decision tree and random forest on large data sets. The model performance is extensively evaluated in terms of different performance measurement parameters such as accuracy, efficiency, scalability and speedup in sequential and distributed environment. The experimental results show that the parallel approach achieved 3.9x computation speedup as compared to the sequential approach without affecting accuracy level. The parallel support vector machine algorithm demonstrated high speedup and scalability in comparison with random forest and decision tree algorithms. The outcome of this study could promote parallel based model for the unobtrusive recognition of lower limb locomotion modes and could promote the future design for the intelligent control of prostheses and exoskeleton.

A Systematic Review of EMG Applications for the Characterization of Forearm and Hand Muscle Activity during Activities of Daily Living: Results, Challenges, and Open Issues.

The role of the hand is crucial for the performance of activities of daily living, thereby ensuring a full and autonomous life. Its motion is controlled by a complex musculoskeletal system of approximately 38 muscles. Therefore, measuring and interpreting the muscle activation signals that drive hand motion is of great importance in many scientific domains, such as neuroscience, rehabilitation, physiotherapy, robotics, prosthetics, and biomechanics. Electromyography (EMG) can be used to carry out the neuromuscular characterization, but it is cumbersome because of the complexity of the musculoskeletal system of the forearm and hand. This paper reviews the main studies in which EMG has been applied to characterize the muscle activity of the forearm and hand during activities of daily living, with special attention to muscle synergies, which are thought to be used by the nervous system to simplify the control of the numerous muscles by actuating them in task-relevant subgroups. The state of the art of the current results are presented, which may help to guide and foster progress in many scientific domains. Furthermore, the most important challenges and open issues are identified in order to achieve a better understanding of human hand behavior, improve rehabilitation protocols, more intuitive control of prostheses, and more realistic biomechanical models.

Contrast of multi‐resolution analysis approach to transhumeral phantom motion decoding

In signal processing, multiresolution decomposition techniques allow for the separation of an acquired signal into sub levels, where the optimal level within the signal minimises redundancy, uncertainties, and contains the information required for the characterisation of the sensed phenomena. In the area of physiological signal processing for prosthesis control, scenarios where a signal decomposition analysis are required: the wavelet decomposition (WD) has been seen to be the favoured time-frequency approach for the decomposition of non-stationary signals. From a research perspective, the WD in certain cases has allowed for a more accurate motion intent decoding process following feature extraction and classification. Despite this, there is yet to be a widespread adaptation of the WD in a practical setting due to perceived computational complexity. Here, for neuromuscular (electromyography) and brainwave (electroencephalography) signals acquired from a transhumeral amputee, a computationally efficient time domain signal decomposition method based on a series of heuristics was applied to process the acquired signals before feature extraction. The results showed an improvement in motion intent decoding prowess for the proposed time-domain-based signal decomposition across four different classifiers for both the neuromuscular and brain wave signals when compared to the WD and the raw signal.

Adaptive Neural Decoder for Prosthetic Hand Control.

The overarching goal was to resolve a major barrier to real-life prosthesis usability—the rapid degradation of prosthesis control systems, which require frequent recalibrations. Specifically, we sought to develop and test a motor decoder that provides (1) highly accurate, real-time movement response, and (2) unprecedented adaptability to dynamic changes in the amputee’s biological state, thereby supporting long-term integrity of control performance with few recalibrations. To achieve that, an adaptive motor decoder was designed to auto-switch between algorithms in real-time. The decoder detects the initial aggregate motoneuron spiking activity from the motor pool, then engages the optimal parameter settings for decoding the motoneuron spiking activity in that particular state. “Clear-box” testing of decoder performance under varied physiological conditions and post-amputation complications was conducted by comparing the movement output of a simulated prosthetic hand as driven by the decoded signal vs. as driven by the actual signal. Pearson’s correlation coefficient and Normalized Root Mean Square Error were used to quantify the accuracy of the decoder’s output. Our results show that the decoder algorithm extracted the features of the intended movement and drove the simulated prosthetic hand accurately with real-time performance (<10 ms) (Pearson’s correlation coefficient >0.98 to >0.99 and Normalized Root Mean Square Error <13–5%). Further, the decoder robustly decoded the spiking activity of multi-speed inputs, inputs generated from reversed motoneuron recruitment, and inputs reflecting substantial biological heterogeneity of motoneuron properties, also in real-time. As the amputee’s neuromodulatory state changes throughout the day and the electrical properties and ratio of slower vs. faster motoneurons shift over time post-amputation, the motor decoder presented here adapts to such changes in real-time and is thus expected to greatly enhance and extend the usability of prostheses.

A pilot exploration on the use of NIR monitored haemodynamics in gesture recognition for transradial prosthesis control

• Evaluation of hemodynamics as a physiological monitor for prosthesis control • Meta-heuristic feature exploration • Pilot to evaluate a potentially cheaper means of prosthesis control • Hand gesture recognition using blood flow dynamics The application of near infrared (NIR) sensing for gesture recognition carries a high appeal as it presents a haemodynamic monitoring method, which provides an alternative to the conventional electromyography sensing method for prosthesis control, which is limited by muscle fatigue and is said to be expensive depending on the specifications. Using an affordable (£25), low sampling and high resolution wearable NIR sensor, a feature selection exercise was conducted where 23 signal features were extracted from the acquired signal and downselected to an optimal 11, which allow for gesture recognition while enhancing recognition accuracy. The selected features were tested and compared with a reduced feature set with four different classifiers, namely multilayer perceptron neural network (MLPNN), Bayesian classifier (BC), linear and quadratic discriminant analysis (LDA and QDA) across four different gestures in five non-amputated participants. The best improvement in classification was produced with the MLPNN and QDA, owing to their overall model complexity and ability to separate data clusters using nonlinear decision boundaries, therein validating the candidate list of features which could be used to characterise and extract more information from haemodynamic based signals. Further work in this area can include an exercise to determine the optimal spacing of the NIR emitters and receivers to allow for maximal penetrative depth and therein an increased amount of physiological information acquired in the signal, and an investigation to observe the extent which anatomical contraction force can be differentiated using NIR.

Evaluation of Continuous Walking Speed Determination Algorithms and Embedded Sensors for a Powered Knee &amp; Ankle Prosthesis

Dynamically altering the parameters for assistance in a lower limb prosthesis is a challenge that depends directly on the ability to estimate gait parameters. Machine learning algorithms present an opportunity to develop methods for continuously determining walking speed in different conditions. Current state-of-the-art solutions involve using wearable sensors such as IMUs to estimate these parameters. These methods require an entire gait cycle to update the walking speed; this leads to delays when responding to changing speeds and ultimately renders these methods ineffective for adaptation into real-time prosthesis control. In this study, we developed subject dependent and independent machine learning models for rapidly determining walking speed and evaluated on data collected from 6 individuals with unilateral transfemoral amputation walking on our robotic knee/ankle prosthesis. We evaluated the performance of these models across a variety of static walking speeds and dynamic speed trials. Our findings suggest that using machine learning models offers excellent accuracy for both subject dependent and subject independent algorithms (DEP RMSE: 0.014 $\pm$ 0.001 m/s, IND RMSE: 0.070 $\pm$ 0.007 m/s, (p < 0.05), with the advantage of real-time continuous determination at 50 Hz, which allows for good performance when rapidly changing walking speed. We also determine the most effective sensors to use for improving model performance. Our study provides valuable information for determining walking speed more reliably across different users and is robust to dynamic changes experienced in gait.

Identification of Gait-Cycle Phases for Prosthesis Control.

The major problem with transfemoral prostheses is their capacity to compensate for the loss of the knee joint. The identification of gait-cycle phases plays an important role in the control of these prostheses. Such control is completely up to the patient in passive prostheses or partly facilitated by the prosthesis in semiactive prostheses. In both cases, the patient recovers his/her walking ability through a suitable rehabilitation procedure that aims at recreating proprioception in the patient. Understanding proprioception passes through the identification of conditions and parameters that make the patient aware of lower-limb body segments’ postures, and the recognition of the current gait-cycle phase/period is the first step of this awareness. Here, a proposal is presented for the identification of the gait-cycle phases/periods under different walking conditions together with a control logic for a possible active/semiactive prosthesis. The proposal is based on the detection of different gait-cycle events as well as on different walking conditions through a load sensor, which is implemented by analyzing the variations in some gait parameters. The validation of the proposed method is done by using gait-cycle data present in the literature. The proposal assumes the prosthesis is equipped with an energy-storing foot without mobility.

Extended home use of an advanced osseointegrated prosthetic arm improves function, performance, and control efficiency

Objective. Full restoration of arm function using a prosthesis remains a grand challenge; however, advances in robotic hardware, surgical interventions, and machine learning are bringing seamless human-machine interfacing closer to reality. Approach. Through extensive data logging over 1 year, we monitored at-home use of the dexterous Modular Prosthetic Limb controlled through pattern recognition of electromyography (EMG) by an individual with a transhumeral amputation, targeted muscle reinnervation, and osseointegration (OI). Main results. Throughout the study, continuous prosthesis usage increased (1% per week, p < 0.001) and functional metrics improved up to 26% on control assessments and 76% on perceived workload evaluations. We observed increases in torque loading on the OI implant (up to 12.5% every month, p < 0.001) and prosthesis control performance (0.5% every month, p < 0.005), indicating enhanced user integration, acceptance, and proficiency. More importantly, the EMG signal magnitude necessary for prosthesis control decreased, up to 34.7% (p < 0.001), over time without degrading performance, demonstrating improved control efficiency with a machine learning-based myoelectric pattern recognition algorithm. The participant controlled the prosthesis up to one month without updating the pattern recognition algorithm. The participant customized prosthesis movements to perform specific tasks, such as individual finger control for piano playing and hand gestures for communication, which likely contributed to continued usage. Significance. This work demonstrates, in a single participant, the functional benefit of unconstrained use of a highly anthropomorphic prosthetic limb over an extended period. While hurdles remain for widespread use, including device reliability, results replication, and technical maturity beyond a prototype, this study offers insight as an example of the impact of advanced prosthesis technology for rehabilitation outside the laboratory.

Joint speed feedback improves myoelectric prosthesis adaptation after perturbed reaches in non amputees

Accurate control of human limbs involves both feedforward and feedback signals. For prosthetic arms, feedforward control is commonly accomplished by recording myoelectric signals from the residual limb to predict the user’s intent, but augmented feedback signals are not explicitly provided in commercial devices. Previous studies have demonstrated inconsistent results when artificial feedback was provided in the presence of vision; some studies showed benefits, while others did not. We hypothesized that negligible benefits in past studies may have been due to artificial feedback with low precision compared to vision, which results in heavy reliance on vision during reaching tasks. Furthermore, we anticipated more reliable benefits from artificial feedback when providing information that vision estimates with high uncertainty (e.g. joint speed). In this study, we test an artificial sensory feedback system providing joint speed information and how it impacts performance and adaptation during a hybrid positional-and-myoelectric ballistic reaching task. We found that overall reaching errors were reduced after perturbed control, but did not significantly improve steady-state reaches. Furthermore, we found that feedback about the joint speed of the myoelectric prosthesis control improved the adaptation rate of biological limb movements, which may have resulted from high prosthesis control noise and strategic overreaching with the positional control and underreaching with the myoelectric control. These results provide insights into the relevant factors influencing the improvements conferred by artificial sensory feedback.

Phantom motion intent decoding for transhumeral prosthesis control with fused neuromuscular and brain wave signals

In recent years, the electroencephalography (EEG) brain–computer interface (BCI) has been researched in the area of upper-limb prosthesis control due to the promise of being able to record neurological signals which follow activation patterns in the cortex directly from the brain with non-invasive electrodes. This is seen as a way of bypassing the limitation posed by acquiring neuromuscular signals predominantly with electromyography (EMG) directly from the stump, which possesses residual limb anatomy post-amputation. In this study, the sequential forward selection algorithm to form a 10-optimal-channel representation, alongside an extended signal feature vector was applied, to investigate the motion intent decoding performance of EMG-only, EEG-only, and a fused EMG–EEG sensing configuration for four transhumeral amputees with varying stump lengths. The results showed a considerable improvement for the EMG-only configuration with the advanced feature vector, but only a small increase for the EEG-only, and thus a marginal improvement when information from both signals was fused together. This is likely due to the EEG requiring a greater number of channels spread across the skull to provide a reliable intent decoding. Further work will now involve optimisation studies to find a greater representation of electrode representation and parsimony, to minimise the number of channels while boosting motion intent decoding accuracy.

Activities of daily living with bionic arm improved by combination training and latching filter in prosthesis control comparison

BackgroundAdvanced prostheses can restore function and improve quality of life for individuals with amputations. Unfortunately, most commercial control strategies do not fully utilize the rich control information from residual nerves and musculature. Continuous decoders can provide more intuitive prosthesis control using multi-channel neural or electromyographic recordings. Three components influence continuous decoder performance: the data used to train the algorithm, the algorithm, and smoothing filters on the algorithm’s output. Individual groups often focus on a single decoder, so very few studies compare different decoders using otherwise similar experimental conditions.MethodsWe completed a two-phase, head-to-head comparison of 12 continuous decoders using activities of daily living. In phase one, we compared two training types and a smoothing filter with three algorithms (modified Kalman filter, multi-layer perceptron, and convolutional neural network) in a clothespin relocation task. We compared training types that included only individual digit and wrist movements vs. combination movements (e.g., simultaneous grasp and wrist flexion). We also compared raw vs. nonlinearly smoothed algorithm outputs. In phase two, we compared the three algorithms in fragile egg, zipping, pouring, and folding tasks using the combination training and smoothing found beneficial in phase one. In both phases, we collected objective, performance-based (e.g., success rate), and subjective, user-focused (e.g., preference) measures.ResultsPhase one showed that combination training improved prosthesis control accuracy and speed, and that the nonlinear smoothing improved accuracy but generally reduced speed. Phase one importantly showed simultaneous movements were used in the task, and that the modified Kalman filter and multi-layer perceptron predicted more simultaneous movements than the convolutional neural network. In phase two, user-focused metrics favored the convolutional neural network and modified Kalman filter, whereas performance-based metrics were generally similar among all algorithms.ConclusionsThese results confirm that state-of-the-art algorithms, whether linear or nonlinear in nature, functionally benefit from training on more complex data and from output smoothing. These studies will be used to select a decoder for a long-term take-home trial with implanted neuromyoelectric devices. Overall, clinical considerations may favor the mKF as it is similar in performance, faster to train, and computationally less expensive than neural networks.