Closed-Loop Multi-Amplitude Control for Robust and Dexterous Performance of Myoelectric Prosthesis.

In the case of a hand amputation, the affected person can use a myoelectric prosthesis to substitute the missing limb and regain motor functions. Unfortunately, commercial methods for myoelectric control, although robust and simple, are unintuitive and cognitively taxing when applied to an advanced multi-functional prosthesis. The state-of-the-art methods developed in academia are based on machine learning and therefore require long training and suffer from a lack of robustness. This work presents a novel closed-loop multi-level amplitude controller (CMAC), which aims at overcoming these drawbacks. The CMAC implements three degrees-of-freedom (DoF) control by thresholding the muscle contraction intensity during wrist flexion and extension movements. Unique features of the controller are the vibrotactile feedback that communicates the state of the controller to the user and a scheme for proportional control. These components allow exploiting the full dexterity of the prosthesis using a simple two-channel myoelectric interface. The CMAC was compared to a commonly implemented pattern-recognition method (linear discriminant analysis - LDA) using clinically relevant tests in 12 able-bodied and 2 amputee subjects. The experimental assessment demonstrated that CMAC was similarly fast as LDA in dexterous tests (clothespin and cube manipulation), while it was somewhat slower than LDA during a simple, single DoF task (box and blocks). In addition, in all the tasks, LDA and CMAC resulted in a similarly low error rate. On the other hand, to an amputee that could not generate six distinguishable classes using LDA, the CMAC still enabled the control of all the prosthesis DoFs. Importantly, the overall setup and training time in CMAC were significantly lower compared to LDA. In conclusion, the novel method is convenient for clinical applications, and allows substantially higher control dexterity compared to what can be normally achieved using conventional two channel EMG. Therefore, CMAC provides performance comparable to advanced machine-learning algorithms and the robustness and ease of use that is characteristic for the simple two-channel myoelectric interface.

Myocontrol is closed-loop control: incidental feedback is sufficient for scaling the prosthesis force in routine grasping.

BACKGROUND: Sensory feedback is critical for grasping in able-bodied subjects. Consequently, closing the loop in upper-limb prosthetics by providing artificial sensory feedback to the amputee is expected to improve the prosthesis utility. Nevertheless, even though amputees rate the prospect of sensory feedback high, its benefits in daily life are still very much debated. We argue that in order to measure the potential functional benefit of artificial sensory feedback, the baseline open-loop performance needs to be established. METHODS: The myoelectric control of naïve able-bodied subjects was evaluated during modulation of electromyographic signals (EMG task), and grasping with a prosthesis (Prosthesis task). The subjects needed to activate the wrist flexor muscles and close the prosthesis to reach a randomly selected target level (routine grasping). To assess the baseline performance, the tasks were performed with a different extent of implicit feedback (proprioception, prosthesis motion and sound). Finally, the prosthesis task was repeated with explicit visual force feedback. The subjects' ability to scale the prosthesis command/force was assessed by testing for a statistically significant increase in the median of the generated commands/forces between neighboring levels. The quality of control was evaluated by computing the median absolute error (MAE) with respect to the target. RESULTS: The subjects could successfully scale their motor commands and generated prosthesis forces across target levels in all tasks, even with the least amount of implicit feedback (only muscle proprioception, EMG task). In addition, the deviation of the generated commands/forces from the target levels decreased with additional feedback. However, the increase in implicit feedback, from proprioception to prosthesis motion and sound, seemed to have a more substantial effect than the final introduction of explicit feedback. Explicit feedback improved the performance mainly at the higher target-force levels. CONCLUSIONS: The study establishes the baseline performance of myoelectric control and prosthesis grasping force. The results demonstrate that even without additional feedback, naïve subjects can effectively modulate force with good accuracy with respect to that achieved when increasing the amount of feedback information.

The clinical relevance of advanced artificial feedback in the control of a multi-functional myoelectric prosthesis.

BACKGROUND: To effectively replace the human hand, a prosthesis should seamlessly respond to user intentions but also convey sensory information back to the user. Restoration of sensory feedback is rated highly by the prosthesis users, and feedback is critical for grasping in able-bodied subjects. Nonetheless, the benefits of feedback in prosthetics are still debated. The lack of consensus is likely due to the complex nature of sensory feedback during prosthesis control, so that its effectiveness depends on multiple factors (e.g., task complexity, user learning). METHODS: We evaluated the impact of these factors with a longitudinal assessment in six amputee subjects, using a clinical setup (socket, embedded control) and a range of tasks (box and blocks, block turn, clothespin and cups relocation). To provide feedback, we have proposed a novel vibrotactile stimulation scheme capable of transmitting multiple variables from a multifunction prosthesis. The subjects wore a bracelet with four by two uniformly placed vibro-tactors providing information on contact, prosthesis state (active function), and grasping force. The subjects also completed a questionnaire for the subjective evaluation of the feedback. RESULTS: The tests demonstrated that feedback was beneficial only in the complex tasks (block turn, clothespin and cups relocation), and that the training had an important, task-dependent impact. In the clothespin relocation and block turn tasks, training allowed the subjects to establish successful feedforward control, and therefore, the feedback became redundant. In the cups relocation task, however, the subjects needed some training to learn how to properly exploit the feedback. The subjective evaluation of the feedback was consistently positive, regardless of the objective benefits. These results underline the multifaceted nature of closed-loop prosthesis control as, depending on the context, the same feedback interface can have different impact on performance. Finally, even if the closed-loop control does not improve the performance, it could be beneficial as it seems to improve the subjective experience. CONCLUSIONS: Therefore, in this study we demonstrate, for the first time, the relevance of an advanced, multi-variable feedback interface for dexterous, multi-functional prosthesis control in a clinically relevant setting.

Simultaneous control of multiple functions of bionic hand prostheses: Performance and robustness in end users.

Myoelectric hand prostheses are usually controlled with two bipolar electrodes located on the flexor and extensor muscles of the residual limb. With clinically established techniques, only one function can be controlled at a time. This is cumbersome and limits the benefit of additional functions offered by modern prostheses. Extensive research has been conducted on more advanced control techniques, but the clinical impact has been limited, mainly due to the lack of reliability in real-world conditions. We implemented a regression-based control approach that allows for simultaneous and proportional control of two degrees of freedom and evaluated it on five prosthetic end users. In the evaluation of tasks mimicking daily life activities, we included factors that limit reliability, such as tests in different arm positions and on different days. The regression approach was robust over multiple days and only slightly affected by changing in the arm position. Additionally, the regression approach outperformed two clinical control approaches in most conditions.

Comparison of fMRI Digit Representations of the Dominant and Non-dominant Hand in the Human Primary Somatosensory Cortex.

The tactile digit representations in the primary somatosensory cortex have so far been mapped for either the left or the right hand. This study localized all ten digit representations in right-handed subjects and compared them within and across the left and right hands to assess potential differences in the functional organization of the digit map between hands and in the structural organization between hemispheres. Functional magnetic resonance imaging of tactile stimulation of each fingertip in BA 3b confirmed the expected lateral-anterior-inferior to medial-posterior-superior succession from thumb to little-finger representation, located in the post-central gyrus opposite to the motor hand knob. While the more functionally related measures, such as the extent and strength of activation as well as the Euclidean distance between neighboring digit representations, showed significant differences between the digits, no side difference was detected: the layout of the functional digit-representation map did not consistently differ between the left, non-dominant, and the right, dominant hand. Comparing the absolute spatial coordinates also revealed a significant difference for the digits, but not between the left and right hand digits. Estimating the individual subject's digit coordinates of one hand by within-subject mirroring of the other-hand digit coordinates across hemispheres yielded a larger estimation error distance than using averaged across-subjects coordinates from within the same hemisphere. However, both methods should only be used with care for single-subject clinical evaluation, as an average estimation error of around 9 mm was observed, being slightly higher than the average distance between neighboring digits.

Longitudinal high-density EMG classification: Case study in a glenohumeral TMR subject.

Targeted muscle reinnervation (TMR) represents a breakthrough interface for prosthetic control in high-level upper-limb amputees. However, clinically, it is still limited to the direct motion-wise control restricted by the number of reinnervation sites. Pattern recognition may overcome this limitation. Previous studies on EMG classification in TMR patients experienced with myocontrol have shown greater accuracy when using high-density (HD) recordings compared to conventional single-channel derivations. This case study investigates the potential of HD-EMG classification longitudinally over a period of 17 months post-surgery in a glenohumeral amputee. Five experimental sessions, separated by approximately 3 months, were performed. They were timed during a standard rehabilitation protocol that included intensive physio- and occupational therapy, myosignal training, and routine use of the final myoprosthesis. The EMG signals recorded by HD-EMG grids were classified into 12 classes. The first sign of EMG activity was observed in the second experimental session. The classification accuracy over 12 classes was 76% in the third session and ∼95% in the last two sessions. When using training and testing sets that were acquired with a 1-h time interval in between, a much lower accuracy (32%, Session 4) was obtained, which improved upon prosthesis usage (Session 5, 67%). The results document the improvement in EMG classification accuracy throughout the TMR-rehabilitation process.

Electrotactile EMG feedback improves the control of prosthesis grasping force.

OBJECTIVE: A drawback of active prostheses is that they detach the subject from the produced forces, thereby preventing direct mechanical feedback. This can be compensated by providing somatosensory feedback to the user through mechanical or electrical stimulation, which in turn may improve the utility, sense of embodiment, and thereby increase the acceptance rate. APPROACH: In this study, we compared a novel approach to closing the loop, namely EMG feedback (emgFB), to classic force feedback (forceFB), using electrotactile interface in a realistic task setup. Eleven intact-bodied subjects and one transradial amputee performed a routine grasping task while receiving emgFB or forceFB. The two feedback types were delivered through the same electrotactile interface, using a mixed spatial/frequency coding to transmit 8 discrete levels of the feedback variable. In emgFB, the stimulation transmitted the amplitude of the processed myoelectric signal generated by the subject (prosthesis input), and in forceFB the generated grasping force (prosthesis output). The task comprised 150 trials of routine grasping at six forces, randomly presented in blocks of five trials (same force). Interquartile range and changes in the absolute error (AE) distribution (magnitude and dispersion) with respect to the target level were used to assess precision and overall performance, respectively. MAIN RESULTS: Relative to forceFB, emgFB significantly improved the precision of myoelectric commands (min/max of the significant levels) for 23%/36% as well as the precision of force control for 12%/32%, in intact-bodied subjects. Also, the magnitude and dispersion of the AE distribution were reduced. The results were similar in the amputee, showing considerable improvements. SIGNIFICANCE: Using emgFB, the subjects therefore decreased the uncertainty of the forward pathway. Since there is a correspondence between the EMG and force, where the former anticipates the latter, the emgFB allowed for predictive control, as the subjects used the feedback to adjust the desired force even before the prosthesis contacted the object. In conclusion, the online emgFB was superior to the classic forceFB in realistic conditions that included electrotactile stimulation, limited feedback resolution (8 levels), cognitive processing delay, and time constraints (fast grasping).

Individual left-hand and right-hand intra-digit representations in human primary somatosensory cortex.

Individual intra-digit somatotopy of all phalanges of the middle and little finger of the right and left hand was studied by functional magnetic resonance imaging in 12 healthy subjects. Phalanges were tactilely stimulated and activation in BA 3b of the human primary somatosensory cortex could be observed for each individual phalanx. Activation peaks were further analysed using the Direction/Order (DiOr) method, which identifies somatotopy, if a significantly high number of subjects exhibit ordered distal-to-proximal phalanx representions along a similar direction. Based on DiOr, ordered and similar-direction-aligned intra-digit maps across subjects were found at the left hand for the little and middle finger and at the right hand for the little finger. In these digits the proximal phalanges were represented more medially along the course of the central sulcus than the distal phalanges. This is contrasted by the intra-digit maps for the middle finger of the right hand, which showed larger inter-subject variations of phalanx alignments without a similar within-digit representation across subjects. As all subjects were right-handed and as the middle finger of the dominant hand probably plays a more individual role in everyday tactile performance than the little finger of the right hand and all left-hand digits, the observed variation might reflect a functional somatotopy based on individual use of that particular digit at the dominant hand.

Individual fMRI maps of all phalanges and digit bases of all fingers in human primary somatosensory cortex.

This study determined the individual maps of all fingers in Brodmann area 3b of the human primary somatosensory cortex in a single fMRI session by tactile stimulation at 19 sites across all phalanges and digit bases of the 5 right-hand digits. To quantify basic features of the digit maps within and across subjects, we applied standard descriptive measures, but also implemented a novel quantitative analysis. This so-called Direction/Order (DiOr) method tested whether subjects exhibited an ordering of peak fMRI representations along their individual direction of alignment through the set of analyzed phalanges and whether these individual directions were similar across subjects. Across-digit analysis demonstrated that for each set of homologous phalanges, the D5-to-D1 representations were successively represented along a common direction of alignment. Hence, the well-known mediolateral D5-to-D1 somatotopy was not only confirmed for the distal phalanges (p1), but could also be shown for the medial (p2) and proximal phalanges (p3). In contrast, the peak activation for the digit bases (p4) only partly elicited that digit succession. Complementary, intra-digit analysis revealed a divergent picture of map topography for the different digits. Within D5 (and in a trend: D4), an ordered p1-to-p3 succession was found across subjects, pointing to a consistent intra-digit somatotopy for D5, with p3 generally found medial-posterior to p1. In contrast, for D1, D2, and D3, most subjects did not present with ordered p1-to-p3 maps nor were directions of alignment similarly oriented between subjects. These digits therefore exhibited highly diverse representation patterns across subjects.

Feature-based attentional modulation of orientation perception in somatosensation.

In a reaction time study of human tactile orientation detection the effects of spatial attention and feature-based attention were investigated. Subjects had to give speeded responses to target orientations (parallel and orthogonal to the finger axis) in a random stream of oblique tactile distractor orientations presented to their index and ring fingers. Before each block of trials, subjects received a tactile cue at one finger. By manipulating the validity of this cue with respect to its location and orientation (feature), we provided an incentive to subjects to attend spatially to the cued location and only there to the cued orientation. Subjects showed quicker responses to parallel compared to orthogonal targets, pointing to an orientation anisotropy in sensory processing. Also, faster reaction times (RTs) were observed in location-matched trials, i.e., when targets appeared on the cued finger, representing a perceptual benefit of spatial attention. Most importantly, RTs were shorter to orientations matching the cue, both at the cued and at the uncued location, documenting a global enhancement of tactile sensation by feature-based attention. This is the first report of a perceptual benefit of feature-based attention outside the spatial focus of attention in somatosensory perception. The similarity to effects of feature-based attention in visual perception supports the notion of matching attentional mechanisms across sensory domains.

Functional MRI indicates consistent intra-digit topographic maps in the little but not the index finger within the human primary somatosensory cortex.

This study explored the question of intra-digit somatotopy of sensory representations in the little and index finger of 10 subjects using tactile stimulation of the fingertip (p1) and base (p4) and functional magnetic resonance imaging (fMRI) at 1.5mm isotropic spatial resolution. The Euclidian distances between p1 and p4 peak representations in Brodmann area 3b resulted in 5.0±0.7mm for the little finger and 6.7±0.5mm for the index finger. These non-collocated representations were found to be consistently ordered across subjects for the little but not the index finger. When using separate distances for medial-lateral, anterior-posterior, and inferior-superior orientations, p4 was 1.9±0.7mm medial to p1 for the little finger in agreement with findings in macaque monkeys, whereas no consistent intra-digit somatotopy across subjects was found for the index finger. This discrepancy could point to differences in the map-forming processes based on sensory input. On the behavioral level it may be attributed to our everyday use of the hand, for which p4 of the index finger plays a much less important role than p4 of the little finger, which is located at the outer border of the hand.