Article

Decoding a new neural-machine interface for control of artificial limbs

Neural Engineering Center for Artificial Limbs, Rehabilitation Institute of Chicago, IL, USA.
Journal of Neurophysiology (Impact Factor: 3.04). 12/2007; 98(5):2974-82. DOI: 10.1152/jn.00178.2007
Source: PubMed

ABSTRACT An analysis of the motor control information content made available with a neural-machine interface (NMI) in four subjects is presented in this study. We have developed a novel NMI-called targeted muscle reinnervation (TMR)-to improve the function of artificial arms for amputees. TMR involves transferring the residual amputated nerves to nonfunctional muscles in amputees. The reinnervated muscles act as biological amplifiers of motor commands in the amputated nerves and the surface electromyogram (EMG) can be used to enhance control of a robotic arm. Although initial clinical success with TMR has been promising, the number of degrees of freedom of the robotic arm that can be controlled has been limited by the number of reinnervated muscle sites. In this study we assess how much control information can be extracted from reinnervated muscles using high-density surface EMG electrode arrays to record surface EMG signals over the reinnervated muscles. We then applied pattern classification techniques to the surface EMG signals. High accuracy was achieved in the classification of 16 intended arm, hand, and finger/thumb movements. Preliminary analyses of the required number of EMG channels and computational demands demonstrate clinical feasibility of these methods. This study indicates that TMR combined with pattern-recognition techniques has the potential to further improve the function of prosthetic limbs. In addition, the results demonstrate that the central motor control system is capable of eliciting complex efferent commands for a missing limb, in the absence of peripheral feedback and without retraining of the pathways involved.

Download full-text

Full-text

Available from: Levi J Hargrove, Jul 08, 2015
1 Follower
 · 
130 Views
  • Source
    • "Feature extraction is a necessary step for pattern recognition. As the effectiveness of time-domain (TD) feature set had been repeatedly demonstrated in previous studies on EMG pattern recognition [7] [24] [25], the TD features were adopted in this study. TD features were originally proposed by Hudgins et al. [7], where continuous EMG signals were segmented into multiple analysis windows and TD features were extracted from each analysis window. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a solution is proposed to predict the finger joint angle using electromyography (EMG) towards application for partial-hand amputees with functional wrist. In the experimental paradigm, the subject was instructed to continuously move one finger (middle finger for able-bodied subjects and index finger for partial-hand amputees) up to the maximum angle of flexion and extension while the wrist was conducting seven different wrist motions. A switching regime, including one linear discriminant analysis (LDA) classifier and fourteen state-space models, was proposed to continuously decode finger joint angles. LDA classifier was used to recognize which static wrist motion that the subject was conducting and choose the corresponding two state-space models for decoding joint angles of the finger with two degrees of freedom (DOFs). The average classification error rate (CER) was 6.18%, demonstrating that these seven static wrist motions along with the continuous movement of the finger could be classified. To improve the classification performance, a preprocessing method, class-wise stationary subspace analysis (cwSSA), was firstly adopted to extract the stationary components from original EMG data. Consequently, the average CER was reduced by 1.82% (p < 0.05). The state-space model was adopted to estimate the finger joint angle from EMG. The average estimation performance (index R2) of the two joint angles of the finger across seven static wrist motions achieved 0.843. This result shows that the finger's joint angles can be continuously estimated well while the wrist was conducting different static motions simultaneously. The average accuracy of seven static wrist motions with and without cwSSA and the average estimation performance of the two joint angles of the finger prove that the proposed switching regime is effective for continuous estimation of the finger joint angles under different static wrist motions from EMG.
    Biomedical Signal Processing and Control 11/2014; 14:265–271. DOI:10.1016/j.bspc.2014.08.004 · 1.53 Impact Factor
  • Source
    • "Studies involving targeted muscle reinnervation (Kuiken, Miller et al. 2005, Kuiken, Li et al. 2009, Bueno, French et al. 2011, Akhtar, Hargrove et al. 2012, Hebert and Lewicke 2012) suggest that simultaneous multi-dimensional control is possible. EMG pattern recognition control algorithms in robotic upper extremity prostheses routinely produce classification rates greater than 95% for multi-dimension joint movement (Khezri and Jahed 2007, Zhou, Lowery et al. 2007, Scheme, Hudgins et al. 2013, Wurth and Hargrove 2013). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Objective. Powered robotic prostheses create a need for natural-feeling user interfaces and robust control schemes. Here, we examined the ability of a nonlinear autoregressive model to continuously map the kinematics of a transtibial prosthesis and electromyographic (EMG) activity recorded within socket to the future estimates of the prosthetic ankle angle in three transtibial amputees. Approach. Model performance was examined across subjects during level treadmill ambulation as a function of the size of the EMG sampling window and the temporal 'prediction' interval between the EMG/kinematic input and the model's estimate of future ankle angle to characterize the trade-off between model error, sampling window and prediction interval. Main results. Across subjects, deviations in the estimated ankle angle from the actual movement were robust to variations in the EMG sampling window and increased systematically with prediction interval. For prediction intervals up to 150 ms, the average error in the model estimate of ankle angle across the gait cycle was less than 6°. EMG contributions to the model prediction varied across subjects but were consistently localized to the transitions to/from single to double limb support and captured variations from the typical ankle kinematics during level walking. Significance. The use of an autoregressive modeling approach to continuously predict joint kinematics using natural residual muscle activity provides opportunities for direct (transparent) control of a prosthetic joint by the user. The model's predictive capability could prove particularly useful for overcoming delays in signal processing and actuation of the prosthesis, providing a more biomimetic ankle response.
    Journal of Neural Engineering 09/2014; 11(5):056027. DOI:10.1088/1741-2560/11/5/056027 · 3.42 Impact Factor
  • Source
    • "Regardless of whether to exploit classification or regression schemes, the first step when developing myoelectrically controlled prostheses is to robustly select the proper number of electrodes to be mounted in a prosthesis socket and the optimal electrode sites. Many previous studies based on pattern-recognition methods have reported that the performance of a myoelectric prosthesis is generally improved with an increasing number of electrodes (Doerschuk et al 1983, Kuruganti et al 1995, Englehart et al 2001, Zhou et al 2007, He et al 2008, Kuiken et al 2009). It was also confirmed for classification based approaches that optimizing electrode positions can further enhance classification accuracy, but this effect becomes weak when using more than four bipolar electrodes (Hargrove et al 2007). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Objective. Recent studies have shown the possibility of simultaneous and proportional control of electrically powered upper-limb prostheses, but there has been little investigation on optimal channel selection. The objective of this study is to find a robust channel selection method and the channel subsets most suitable for simultaneous and proportional myoelectric prosthesis control of multiple degrees-of-freedom (DoFs). Approach. Ten able-bodied subjects and one person with congenital upper-limb deficiency took part in this study, and performed wrist movements with various combinations of two DoFs (flexion/extension and radial/ulnar deviation). During the experiment, high density electromyographic (EMG) signals and the actual wrist angles were recorded with an 8 × 24 electrode array and a motion tracking system, respectively. The wrist angles were estimated from EMG features with ridge regression using the subsets of channels chosen by three different channel selection methods: (1) least absolute shrinkage and selection operator (LASSO), (2) sequential feature selection (SFS), and (3) uniform selection (UNI). Main results. SFS generally showed higher estimation accuracy than LASSO and UNI, but LASSO always outperformed SFS in terms of robustness, such as noise addition, channel shift and training data reduction. It was also confirmed that about 95% of the original performance obtained using all channels can be retained with only 12 bipolar channels individually selected by LASSO and SFS. Significance. From the analysis results, it can be concluded that LASSO is a promising channel selection method for accurate simultaneous and proportional prosthesis control. We expect that our results will provide a useful guideline to select optimal channel subsets when developing clinical myoelectric prosthesis control systems based on continuous movements with multiple DoFs.
    Journal of Neural Engineering 08/2014; 11(5):056008. DOI:10.1088/1741-2560/11/5/056008 · 3.42 Impact Factor