Subjects
Three subjects with transhumeral amputation where recruited to participate in this study (hereafter referred to as S1, S2 and S3). Subjects S1 and S2 were implanted with e-OPRA in 2017, whereas S3 was implanted previously in 2013. Additionally, S1 and S2 underwent a Targeted Muscle Reinnervation (TMR) surgical procedure for redirecting the radial nerve into the lateral head of the triceps brachii muscle, and the ulnar nerve into the short head of the biceps brachii muscle, aiming for intuitive myoelectric signals for hand open and close, respectively [21, 22]. Bipolar epimysial electrodes were implanted on these TMR reconstructions as well as on the naturally innervated long head of the triceps brachii muscle and on the long head of the biceps brachii muscles. The implanted electrodes were accessed via the e-OPRA Implant System. The tests were conducted between February 2018 and May 2018. The study was approved by the Swedish regional ethical committee in Gothenburg (Dnr: 769–12).
Materials
The subjects performed the VET and PLT while operating their daily transhumeral prostheses that were composed of a myoelectric-locking elbow and a myoelectric terminal device (12 K50 Elbow and VaryPlus Hand, Ottobock, Germany). The terminal device was controlled using the conventional direct (one-for-one) and proportional control strategy [23], fed by either sEMG or eEMG signals. Concerning the sEMG configuration, the two muscle sites were optimally targeted with MyoBock electrodes (Ottobock, Germany) connected directly to the terminal device. This configuration represents the most common solution for myoelectric control in clinical practice. The eEMG configuration used the signals from epimysial electrodes amplified and filtered by a custom-designed embedded system contained within the prosthetic device [24]. The algorithm implemented for the one-for-one control resembled the behavior of conventional MyoBock electrodes. The implementation was based on the direct mapping of the speed of each prosthetic hand movement to the Mean Absolute Value of its corresponding channel, calculated from 50 ms non-overlapping windows of eEMG data sampled at 500 Hz [25, 26].
For each subject, the muscles used for control were the same for both sEMG and eEMG configurations. In the case of S1 and S2, the TMR reconstructions on triceps and biceps muscles were voluntary contracted by respectively opening and closing the phantom hand, these contractions were then directly mapped to the corresponding movement of the prosthetic hand. For S3, the contractions of the naturally innervated portion of triceps and biceps were directly mapped to open and close the prosthesis, as this subject has done daily for the last decade. The prosthesis was, for both control configurations, mechanically attached to the subject’s stump via a clamp mechanism over the percutaneous portion of the osseointegrated implant.
Experimental protocol
The subjects performed the VET and PLT with either sEMG or eEMG configurations (Fig. 1). Both tests were performed while seated in front of a table. The chair height and its distance from the table were adjusted for each subject to achieve a comfortable position to interact with the test objects. The prosthetic elbow joint was kept in a fully extended position during the entire experiment.
The VET was used to measure the subjects’ ability to regulate grip force, which is also affected by the reliability of the interface, while maneuvering the prosthesis in the context of delicate grasping (i.e. when handling fragile objects). The VET was first presented by Clemente et al. as a modification of the well-known box and blocks test for gross manual dexterity [14, 27], and resembles a task of picking and repositioning fragile objects without breaking them. Here, 50 × 50 × 50 mm3 plastic blocks weighing 55 g were equipped with a magnetic fuse, using a magnetic latching mechanism placed in between the opposite walls of the block. A force applied on the walls exceeding a fixed threshold caused the fuse to break instantaneously, similarly to “breaking an egg”. As the VET prescribes, the subjects were asked to transfer the blocks from one side of a plastic wall (height of 15 cm) to the other, as quickly as possible within one minute while also preventing their breakage. The subjects were instructed to complete the action of transferring a block even if this broke while grasping it or during transfer. The number of broken blocks and the total number of transferred blocks (thus comprising both the blocks that remained intact and that broke) were measured. The subjects were asked to perform ten one-minute sessions, half of which were performed with blocks having thresholds set at 18 ± 0.2 N (mean ± sd, VET18N) and half of which with blocks having thresholds set at 6 ± 0.2 N (VET6N). Before the evaluation, all subjects performed a single training session to become accustomed with the task.
Picking and lifting an instrumented object was used to assess the subjects’ motor coordination as well as the reliability of the interface when maneuvering the prosthesis in the context of routine grasping (PLT) [15, 28]. The test object consisted of a 40 × 45 × 130 mm3 plastic block (~ 200 g) with three embedded load cells (SMD2551, Strain Measurements Devices, UK), of which two measured the grip force of the thumb and fingers independently and the third measured the load force applied on the object before lift-off. The test consisted of five series of 20 repetitions each (100 repetitions in total), performed in a single experimental session. Each repetition was performed at self-selected speed and consisted of: 1) moving the arm to reach the object, 2) grasping the object, 3) lifting the object a few centimeters above the desk, 4) repositioning the object back on the table and 5) releasing the object. The grasped surfaces of the test object were covered with sandpaper. The coefficient of friction between the object and the prosthesis was found to be 0.9. Consequently, the minimum grip force required to lift the object was 1.9 N, corresponding to a minimum grip/load force ratio (i.e. the slip GLFr) of 1.1. A repetition was deemed successful if it was performed without overcoming a predefined limit on grip force, set to 30 N (GFTHRESH –software break signaled to the subjects by audio-visual feedback), corresponding to a GLFr of 16. The grip force limit was set empirically after pilot experiments. The limit was introduced to prevent the subjects adopting a strategy in which they grasped the object by fully closing the prosthetic digits. Each subject performed a short training session comprising 10 repetitions.
Data analysis and statistical methods
In the VET, the number of broken and transferred blocks were compared with regards to control configurations (sEMG and eEMG) and used as an assessment for grip force control. In particular, we computed the probability of the observed occasions in which the number of broken or transferred blocks in the eEMG configuration was different than in the sEMG configuration for all subjects assuming a binominal distribution (i.e., B(x; n, p) with x equal to the number of successes, n equal to 3 corresponding to the number of subjects in the group, and p = 0.5 (assuming that the two events have the same probability to happen), akin to Clemente et al. [14].
In the PLT, motor coordination was evaluated through the level of grip-load force (GF-LF) coordination, quantified through the temporal delay between the instants when the GF and LF reached 50% of the LF at lift-off. In addition, we calculated two other performance metrics from the PLT, namely the maximum grip force during a repetition (GFMAX), and the difference between the grip force at lift-off and GFMAX during the holding phase (ΔGF). Specifically, the GFMAX provided an indication of the subjects’ grasp force instinctively produced at lift-off, whereas the ΔGF was related to potentially involuntary changes in the grip force during hold (i.e., owing to cross-talk or motion artifacts). As such, lower GFMAX and ΔGF values are considered to indicate better performance. These metrics provide insights about both grip force control and reliability of the interface, two aspects that should be considered together as a trustworthy assessment of control is only possible when a reliable interface is available. The Wilcoxon rank-sum test was used to compare these performance metrics across configurations.
In all cases, a p value lower than 0.05 was considered as reference for statistical significance. We would like to point out that these commonly used statistical methods were used to analyze the results of the study for scientific rigor. Albeit useful, in this case the information on statistical significance is limited because of the limited number of subjects involved in the study. The restricted availability of subjects recipient of the e-OPRA Implant System limited our possibility to carry out a larger study. For these reasons, in the following we present and discuss the results in a descriptive fashion.