Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Experimental Implementation of Automatic Control of Posture-Dependent Stimulation in an Implanted Standing Neuroprosthesis

Experimental Implementation of Automatic Control of Posture-Dependent Stimulation in an Implanted... Hindawi Applied Bionics and Biomechanics Volume 2019, Article ID 2639271, 11 pages https://doi.org/10.1155/2019/2639271 Research Article Experimental Implementation of Automatic Control of Posture-Dependent Stimulation in an Implanted Standing Neuroprosthesis 1,2 2 1,2 1,2 Brooke M. Odle , Lisa M. Lombardo , Musa L. Audu , and Ronald J. Triolo Department of Biomedical Engineering, Case Western Reserve University, Cleveland 44106, USA Motion Study Laboratory, Louis Stokes Cleveland Veterans Affairs Medical Center, Cleveland 44106, USA Correspondence should be addressed to Brooke M. Odle; brooke.odle@case.edu Received 23 May 2018; Accepted 13 January 2019; Published 14 March 2019 Academic Editor: Le Ping Li Copyright © 2019 Brooke M. Odle et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Knowledge of the upper extremity (UE) effort exerted under real-world conditions is important for understanding how persons with motor or sensory disorders perform the postural shifts necessary to complete many activities of daily living while standing. To this end, a feedback controller, named the “Posture Follower Controller”, was developed to aid in task-dependent posture shifting by individuals with spinal cord injury standing with functional neuromuscular stimulation. In this experimental feasibility study, the controller modulated activation to the paralyzed lower extremity muscles as a function of the position of overall center of pressure (CoP), which was prescribed to move in a straight line in forward and diagonal directions. Posture-dependent control of stimulation enabled leaning movements that translated the CoP up to 48 mm away from the nominal position during quiet standing. The mean 95% prediction ellipse area, a measure of the CoP dispersion in the forward, 2 2 2 forward-right, and forward-left directions, was 951 0 ± 341 1mm , 1095 9 ± 251 2mm , and 1364 5 ± 688 2mm , respectively. The average width of the prediction ellipses across the three directions was 15.1 mm, indicating that the CoP deviated from the prescribed path as task-dependent postures were assumed. The average maximal UE effort required to adjust posture across all leaning directions was 24.1% body weight, which is only slightly more than twice of what is required to maintain balance in an erect standing posture. These preliminary findings suggest that stimulation can be modulated to effectively assume user-specified, task-dependent leaning postures characterized by the CoP shifts that deviate away from the nominal position and which require moderate UE effort to execute. 1. Introduction device, such as a walker or a countertop. To address this lim- itation, previous groups explored closed-loop feedback con- trol systems for standing with stimulation employed at Spinal cord injury (SCI) often results in partial or total individual joints [3–8] as well as a stimulation controller paralysis of the trunk and lower extremity (LE) muscles. Implanted neuroprostheses (NPs) utilizing functional neuro- based on comprehensive or global joint feedback combined with center of mass (CoM) acceleration that rejected destabi- muscular stimulation (FNS) can restore basic standing func- lizing perturbations and reduced the UE effort to maintain tion in individuals with SCI, providing them with the standing balance [9]. However, these advanced control sys- independence to accomplish several activities of daily living tems have been designed to maintain only a single upright [1, 2]. Standing NPs supply constant preprogrammed open- loop stimulation to the trunk, hip, and knee extensors to setpoint in the nominal standing position. Users are only able to stand optimally and resist potentially destabilizing pertur- maintain a single, upright stance. Thus, to maintain balance bations in one erect, neutral posture rather than at forward- in the presence of postural perturbations, NP users rely on or side-leaning postures best suited for specific functional voluntary upper extremity (UE) effort exerted on a support 2 Applied Bionics and Biomechanics The aims of this work are (1) to explore the effectiveness tasks [10, 11]. To adjust posture away from the erect stance with existing systems that deliver preprogrammed patterns of a PFC to enable NP users to lean away from the erect pos- of stimulation, users must exert voluntary UE effort to push ture and (2) to examine the contribution of UE effort to lean- ing postures in real subjects with SCI using a NP for standing. or pull against the support device. At these new positions, the patterns of stimulation tuned for erect standing become In this exploratory study, standing performance was deter- suboptimal and may over or understimulate the muscles mined by the following metrics: (1) maximum resultant UE required to maintain the new task-dependent postures. effort contributed to leaning movements in the AP and ML One benefit to the ability to adjust posture from the erect directions as LE muscle activation is modulated and (2) CoP-tracking deviations from a prescribed straight-line path. stance is the ability to prepare for a functional task, such as reaching and manipulating objects on shelves. This also gives users the ability to reach the full extent of their standing 2. Materials and Methods workspace, thereby providing them with greater indepen- dence and access to objects in the environment. Another ben- 2.1. Subject and Standing Neuroprosthesis System. A 27-year- efit is adjusting posture laterally, to rest muscles on one side old male with motor incomplete C5 tetraplegia (AIS C) par- of the body as a means of mitigating fatigue and prolonging ticipated in the experiments. He was approximately 185.4 cm overall standing times. These benefits are further supported tall and weighed 58.5 kilograms when the experiments were by the work of Abbas and Gillette [12], which suggested that conducted. He received a 16-channel implanted LE NP one shifting posture in the anterior-posterior (AP) and medial- year prior to data collection and was a regular user for recon- lateral (ML) directions would be critical in enabling standing ditioning exercise and standing. At the time of the study, he NP users to accomplish various functional tasks. could stand quietly in the neutral position for 25 minutes To achieve these changes in posture, the location of the with 93% body weight (BW) supported by his legs, utilizing projection of the total body CoM on the base of support will UE effort only for a light touch on the walker to maintain bal- have to change smoothly and continuously. Thus, control ance. Prior to participating in the experiments, the subject systems that can automatically maintain standing balance signed informed consent forms approved by the Institutional should include CoM position feedback and modulate Review Board of the Louis Stokes Cleveland Veterans Affairs stimulation to the LE muscles as posture is adjusted away Medical Center. from the erect stance. One such system (the Posture Follower The standing NP consisted of one surgically implanted Conftroller or PFC) was developed and tested in simulation 16-channel stimulator telemeter [13]. The implanted system [10]. In that simulation feasibility study, the model exerted targeted the following muscle groups: hip extensors (Gluteus voluntary UE effort to manually adjust the CoM location Maximus (GMX), Hamstring (HM), and posterior portion of from an initial neutral setpoint towards the new desired for- the Adductor Magnus (PA)), hip abductors (left Gluteus ward- or side-leaning posture, while the controller continu- Medius (GMED)), trunk extensors (Lumbar Erector Spinae ally updated the neural stimulation to maintain activations (ES)), and the trunk lateral benders (Quadratus Lumborum that were optimal for each change in position. Controller per- (QL)). A selective, multicontact, flat interface nerve cuff elec- formance was measured with respect to (1) the UE effort, trode [14] was implanted on the proximal femoral nerve near defined as the UE forces exerted as the simulated user chan- the inguinal ligament, to activate the three uniarticular vasti ged postures and (2) the ability to track a moving object with muscles of the quadriceps (QD) while avoiding recruitment the CoM as the object moved in the forward, diagonal, and of the sartorius and the biarticulated rectus femoris, which lateral directions. The PFC reduced UE effort by an average induce hip flexion that compromises erect neutral standing. of 50%, compared to using the UEs alone. In general, CoM All other muscles were activated with surgically implanted tracking with the PFC and UE alone was similar, except for intramuscular electrodes [15]. Pulse amplitudes (0.8, 2.1, one instance of an overshoot when moving in the left- 18, or 20 mA) were set on a channel-by-channel basis while diagonal direction without the active controller. These pulse duration (0-250 μs) and frequency (0-20 Hz) were encouraging simulation findings were encouraging and sup- modulated independently on a pulse-by-pulse basis on each port the development of posture-dependent control of stim- channel to achieve the desired motion. A rechargeable ulation for a standing NP. wearable external control unit (ECU) [16] delivered power The next step in the development and deployment of and command signals to the implanted pulse generator via posture-dependent control systems is to implement the a close-coupled inductive link maintained by a transmitting PFC in the laboratory environment with a standing NP user. coil taped to the skin over the stimulator. The ECU coordi- The simulation results in [10] used an ideal UE controller nated the delivery of temporal patterns of stimulation that modulated UE forces as a function of shoulder displace- through all 16 channels simultaneously. ment and velocity away from the erect posture position. The To target the right GMED for postural control in the ML actual UE forces exerted by a standing NP user are generally direction and bilateral tibialis anterior (TA) and gastrocne- obtained from measuring devices in the laboratory and could mius (GS) for postural control in the AP direction, the vary considerably between individuals. Knowledge of the implanted stimulation system was supplemented with self- actual UE effort required by standing NP users to change adhesive surface electrodes. These muscles were recruited posture will greatly help in the design of control systems for because they were not available in the subject’s implanted ensuring balance during performance of important activities system. Surface stimulation was delivered at a constant fre- of daily living. quency of 20 Hz, variable pulse width up to 250 μs dependent Applied Bionics and Biomechanics 3 on controller output, and a fixed pulse amplitude (100 mA where CoP and CoP are the AP and ML components of AP ML for the right GMED and bilateral GS, and 30 mA for the bilat- the CoP. CoP , CoP , CoP , and CoP are the x and y Lx Ly Rx Ry eral TA). components of the CoP, and F and F are the vertical reac- L R Real-time control of stimulation was implemented with a tion forces under the left and right feet, respectively. custom software developed in MATLAB/Simulink R7.9 and As the subject leans away from the erect stance and the the xPC Target toolbox (MathWorks Inc., Natick, MA). A CoP moves away from the nominal (erect) position, the Windows (Microsoft Inc., Redmond, WA) host computer resulting changes in AP and ML components are tracked was utilized to build customized applications, while a dedi- by the PFC via a simple proportional control law. Assuming cated (target) computer with the Pentium Dual-Core 3 GHz a linear relationship between changes in the CoP and muscle microprocessor (Intel Inc., Santa Clara, CA) with 2 GB of activation, the changes in activation to be applied to the LE muscles are computed according to (equation 3): RAM was responsible for running the applications in real time. The host and target computers communicated via the TCP/IP protocol. Data were acquired using a NI PCI-6071E SPW‐BPW ∗ ΔCoP Muscle activation = BPW + Gain ∗ , board (National Instruments Inc., Austin, TX). For the Max CoP iD experiments described, all real-time controller and stimula- tion parameters were sampled at 40 Hz. The stimulus values for erect standing were determined by clinical observation whereby the subject exhibited ample knee, hip, and trunk where BPW is the baseline pulse width value, Gain is the extension to achieve an erect posture without discomfort. proportional gain applied to the AP and ML components of Baseline standing stimulation values are listed in Table 1. the directions of the overall CoP, SPW is the muscle satura- tion pulse width value (defined as the maximum above which 2.2. Posture Follower Controller Design. In this study, the no additional force is generated), CoP is the subject’s instan- overall center of pressure (CoP) position (a function of the taneous CoP position, and Max CoP is the maximum iD location of the vertical ground reaction force vector) was excursion of the path of the object being tracked by the sub- used as the feedback signal for the PFC. A more suitable feed- ject’s CoP relative to the nominal position. In this equation, i back signal would be the orthogonal projection of the is a placeholder for either the AP or ML component of CoP. whole-body CoM or center of gravity (CoG). However, there Thus, ΔCoP is the difference between the current CoP posi- are challenges in implementing the CoM position (a function tion and the nominal CoP position in the AP or ML direc- of the location of the total body mass, as the feedback signal). tion. The gain setting was limited to values between 0 and Currently, there is no means for the quantity to be computed 1, to prevent the muscle activation from exceeding SPW. or estimated from body-mounted sensors in real time rapidly Assuming that posture is adjusted in a slow and and accurately enough to use as a stimulus control signal quasi-static manner, the PFC targets muscles to provide sup- with a paralyzed user. The CoM, CoG, and CoP are equiva- port, supplying stimulation that is optimal, as determined in lent during static conditions. Thus, the overall CoP position [10], for the static position at any time. Previous simulation was used as a surrogate because it can be readily obtained studies [10, 18] determined that the bilateral GS, GMED, from two force plates (AMTI, Watertown, MA) in the labora- and PA provide support as the postural shifts are elicited tory, making it a more practical control signal for this explor- away from the erect stance. Based on these findings, the atory study. The laboratory-based PFC took the form of a PFC only modulates activation to those muscles. The GS proportional feedback controller, so it tracks voluntary are ankle plantar flexors and were modulated as posture changes in posture by mapping changes in CoP to changes was adjusted forward and backward in the sagittal plane. in LE muscle activations (Figure 1). The TAs are ankle dorsiflexors and were activated at a fixed The user stood at an erect, nominal stance with baseline pulse width to cocontract with the GS, to increase ankle joint (open-loop) stimulation. In this stance, the user stood upright stiffness or, as we observed in the current subject, to help mit- with the feet approximately under the shoulders and each on a igate any spasms that might be triggered as posture is separate force plate. The erect, nominal stance is biomechan- adjusted. The GMED and PA are hip ab/adductors, respec- ically defined as the standing posture in which the head, tively, and were utilized to effect postural shifts in the ML trunk, pelvis, and LEs are aligned as close to vertical as possi- direction. When posture was adjusted leftward, the right ble in sagittal and coronal planes with minimal to no axial GMED and left PA were targeted for activation to support rotation in the coronal plane. The components of the overall the body against the pull of gravity. Conversely, when pos- CoP position in the AP and ML direction were computed ture was adjusted rightward, the left GMED and right PA using (equation (1)) and (equation (2)), respectively [17]. were targeted for activation. The PFC did not modulate mus- cle activation to the extensor muscles (right HM, right ES, F F L R CoP = CoP ∗ + CoP ∗ , 1 and bilateral GMX and QD) which had to be maximally stim- AP Lx Rx F +F F +F L R L R ulated to maintain an erect standing posture. 2.3. Visual Feedback. To assess the standing performance F F L R CoP =CoP ∗ + CoP ∗ , ML Ly Ry with respect to maintaining and tracking posture according F +F F +F L R L R to a prescribed path, visual feedback of the overall CoP and specified paths were presented on a computer monitor in real 4 Applied Bionics and Biomechanics Table 1: Muscle pulse amplitudes and pulse widths for baseline standing. Muscles that were always recruited for baseline standing are indicated with a “P”, while those recruited by the controller are indicated with a “C.” Muscles that were supplemented with surface electrodes are indicated with a “ ”. Baseline standing Muscle Function Pulse amplitude (mA) Threshold PW (μs) Saturation PW (μs) PW (μs) Right gluteus maximus (right GMX) P 20.0 248 2 250 Right hamstring (right HM) P 20.0 250 64 250 Left gluteus maximus (left GMX) P 20.0 145 5 150 Left gluteus medius (left GMED) C 20.0 61.5 13 110 Right quadratus lumborum (right QL) —— — — — Right erector spinae (right ES) P 2.1 90 10 125 Left quadratus lumborum (left QL) — 18.0 0 10 50 Left erector spinae (left ES) — 18.0 0 35 70 Right quadriceps 1 (right QD 1) P 0.8 90 30 90 Right quadriceps 2 (right QD 2) P 0.8 90 24 90 Left quadriceps 1 (left QD 1) P 0.8 250 64 250 Left quadriceps 2 (left QD 2) P 0.8 250 48 250 Right quadriceps 3 (right QD 3) P 0.8 100 32 100 Right posterior adductor (right PA) C 20.0 86 2 170 Left quadriceps 3 (left QD 3) P 0.8 250 72 250 Left posterior adductor (left PA) C 20.0 128.5 7 250 Right gluteus medius (right GMED) 100.0 0 80 250 Right tibialis anterior (right TA) 30.0 0 80 100 Right gastrocnemius (right GS) 100.0 0 20 65 Left tibialis anterior (left TA) C 30.0 0 70 90 Left gastrocnemius (left GS) 100.0 0 30 70 CoP to muscle User’s Muscle activation Erect stance lower activation conversion extremities (Nominal CoP position) − (posture follower controller) Actual CoP position (CoP AP, CoP ML) Figure 1: Control setup. The user stands erect on force plates, which measure the center of pressure (CoP) position in the anterior-posterior (AP) and medial-lateral (ML) directions. The user leans away from the erect stance, adjusting the overall CoP position towards the ends of the paths in the forward and diagonal directions. The force plates continuously track the resulting changes in the CoP position and the posture follower controller converts the changes in the CoP position to muscle activation, which is applied to the lower extremities. time (Figure 2). To ensure that posture was adjusted at a con- 40 mm and 48 mm from the nominal starting position in the forward and diagonal directions, respectively. sistent rate, the subject adjusted his CoP to track a circle moving along a straight line on the computer screen. The moving circle traveled at a speed of 20 mm/s to the end of 2.4. Posture Follower Controller Tuning. The proportional the selected path and returned to the nominal position along gain settings in (equation 3) were tuned by hand over several the same path. This speed was selected because it was the experimental sessions. To determine the optimal settings, the subject stood erect on the force plates with visual feedback maximum speed that enabled the subject to adjust CoP in a continuous manner. The length of each of the paths was and adjusted posture to place his CoP at the location of the based on the subject’s comfort while leaning forward and endpoints of the paths in the forward and diagonal direc- diagonally. The endpoints of each path were positioned tions. If the gain settings were too high, activation to the Applied Bionics and Biomechanics 5 End of forward (FO) path End of forward-right (FR) path (end of current specified path) Current specified path End of forward- le (FL) path Moving circle (tracking signal) Subject’s CoP Nominal (NO) starting position Figure 2: Diagram of the visual feedback display. The subject stood erect at the nominal (NO) starting position and adjusted the overall center of pressure (CoP) position to track the moving circle to the end of the paths defined by the yellow circles. The subject tracked the moving circle along the same path to return to the NO position. Prior to conducting the experiments, the speed of the moving circle and the locations of the endpoints of the paths were tuned to ensure that the subject adjusted posture at a comfortable rate and within reasonable limits of his standing balance. During the experiments, the subject adjusted the overall CoP position (green) to track the moving circle (blue) in the forward (FO), forward-right (FR), and forward-left (FL) directions. As an additional visual cue, the currently specified path is defined by changing the color of its endpoint from yellow to red. In this image, the currently specified path is the one from NO to FR. targeted muscles may require the subject to exert higher UE 2.6. Data Analysis. A repetition is distinguished by leaning forces to resist the muscle actions. Conversely, if the gain set- movement onset and offset (Figure 4). Movement onset was tings were too low, the controller would only nominally mod- defined as the initial time point where the moving circle departed from the nominal starting position. Upon reaching ulate activation to the LEs and diminish its potential impact on the posture. The gain values that enabled the subject to the end of the path, the moving circle dwelled there for 3 sec- comfortably adjust posture to and from the ends of all the onds before returning to the nominal starting position. paths were implemented in all the subsequent repetitions Movement offset was defined as the time point where the with the moving circle. moving circle first acquired the nominal starting position on its return. 2.5. Data Capture. The setup for testing the effects of the PFC There were twelve repetitions in which the subject con- is depicted in Figure 3. After activating baseline stimulation sistently maintained the starting position before movement to transition from sitting to upright standing, the subject onset, tracked the moving circle the entire distance to the donned a suspension harness (McMaster-Carr Inc., Elm- end, and maintained the same nominal starting position hurst, IL) attached to a lanyard (Guardian Fall Protection, after movement offset. Those repetitions were selected for Kent, WA) connected to a hook bolted into the laboratory analysis, and the overall CoP profiles were computed. To ceiling decking for safety. The subject stood with his hands obtain the changes in the CoP relative to the value at the nominal position, the starting CoP position was subtracted on a custom-built adjustable-instrumented walker (80/20, Columbia City, IN), which was adjusted for his height and from the resulting trajectories. UE effort, defined as the comfort. The subject also stood with each foot placed on a maximum resultant UE forces exerted with the PFC, was separate force platform to compute the overall CoP posi- compared to the values exerted during the erect stance tion in the AP and ML directions (equations 1 and 2). (equation 4). Upon settling into a comfortable erect standing position, the locations of the subject’s feet were marked with a tape UE − UE NO D Percent difference = to ensure the same foot placement throughout the experi- ∗ 100, 4 UE NO mental session. A static trial was collected to obtain the UE forces exerted on the walker at the nominal erect standing posture. where UE is the maximum UE force exerted at the nom- NO After instruction and sufficient practice to obviate learning inal position, and UE is the mean maximal UE effort effects, the subject adjusted posture by exerting volitional exerted while changing posture along the three directions UE effort on the instrumented walker to ensure his overall (FO, FR, and FL). Given the quasi-static nature of the CoP tracked the moving circle as it moved in the forward tracking tasks performed in this feasibility study, the 95% (FO) and diagonal directions (forward-right, FR; forwar- prediction ellipse area (PEA) was computed to describe d-left, FL). Five trials were collected, with two repetitions the dispersion of the CoP position in the directions investi- for tracking the circle to the ends of each of the three paths gated. The 95% prediction ellipse represents a region that and returning to the nominal erect position completed per contains the center of the points of the postural sway with trial. The sequence of directions was randomized to avoid 95% probability. Schubert and Kirchner [19] recommended systematic error. the PEA as a standard method of measuring posturographical 6 Applied Bionics and Biomechanics VICON cameras Safety lanyard and harness Computer monitor (visual feedback display) Reflective markers Instrumented walker Force plates Figure 3: Set-up for experimental evaluation of the controller. The subject stands erect on force plates, while holding onto an instrumented walker and adjusting the overall center of pressure (CoP) position towards the end of paths in the forward and diagonal directions. The subject was provided with visual feedback while adjusting the overall CoP position. Reflective markers were mounted on the subject to track his joint positions as he adjusted posture. CoP profiles during tracking task Dwell Dwell Movement Movement onset onset Movement Movement offset offset 105 110 115 120 125 130 Dwell Dwell Movement Movement Movement Movement onset onset offset offset 105 110 115 120 125 130 Trial time (secs) Subject Moving target Figure 4: The sample CoP profiles of the subject (blue) and the moving circle (red) of the two consecutive leaning movements during the tracking task. The top panel displays the CoP profiles in the AP direction, while the bottom panel displays the CoP profiles in the ML direction. Each movement is considered a separate repetition, which consists of a movement onset, dwell period, and movement offset. Movement onsets are indicated as the time point in which the moving circle initiates movement from the nominal starting position to the end of the specified path. Upon reaching the end of the specified path, the moving circle dwells there for 3 seconds. When the dwell period ends, the moving circle returns to the nominal position and remains there until it initiates travel along the next path. The first time point at which the moving circle acquires the nominal starting position on the return is the movement offset. Based on the orientation of the laboratory coordinate system, postural adjustments in the forward direction are indicated as CoP increasing from the nominal AP starting position. The postural shifts towards the left are indicated as CoP increasing from the nominal. Thus, in both repetitions, the ML subject was tracking the moving circle in the forward-left direction. CoP ML (mm) CoP AP (mm) Applied Bionics and Biomechanics 7 maximum resultant UE forces exerted while changing pos- scatter data, instead of the confidence ellipse. PEA was computed as shown in (equation 5). ture were computed for each leaning direction and normal- ized as percentage of BW (Figure 6). The mean maximum resultant UE effort exerted while eliciting leaning movements PEA = π ∗ χ ∗ det S , 5 in the FO, FR, and FL directions were 22 5±0 9%BW, 14 6 ±4 1%BW, and 35 2± 1 3%BW, respectively. As a reference, where χ is the inverse of the chi-square cumulative distribu- the maximum resultant UE force (6.75% BW) exerted while tion function with 2 degrees of freedom at a fixed probability standing in the nominal (NO) starting position is also dis- level (P = 95%, χ ≈ 5 99146), det(S) is the determinant 20 95,2 played. Compared to the maximum resultant UE force of the Eigenvalues of the sample variance covariance matrix exerted at NO, the percent difference in the mean maximum of CoP and CoP . The PEA and the width of each predic- resultant UE effort exerted during leaning movements in AP ML tion ellipse, a measure of CoP deviation from specified path, the FO, FR, and FL directions was 233.3%, 116.3%, and quantified the CoP-tracking performance. 421.5%, respectively. CoP excursions in the AP and ML directions during the 3. Results CoP-tracking task are displayed in Figure 7(a) and are repre- sentative of one repetition of leaning movements in each 3.1. Controller Tuning. When movements were elicited in the direction. To adjust posture in the FO direction, changes in forward or diagonal directions at gain settings larger than 0.4 CoP position in the AP direction were mainly required, with of the changes in CoP , the modulated activation to the AP minimal changes to CoP in the ML direction (as also indi- bilateral GS resulted in raising the heels off the ground so cated in Figure 5(a)). To lean in the FR direction, posture the subject stood on his toes. This heel-raising effect was was adjusted about 10 mm more in the ML direction than diminished when the gain was set to values below 0.35 and the AP direction. While leaning in the FL direction, posture the SPWs of the right and left GS were reduced from 100 μs was adjusted about the same distance in the AP and ML to 65 μs and from 90 μsto 70 μs, respectively. The tuned directions. However, maintaining posture at the end of each SPW for the target muscles are listed in Table 1. When the path required adjusting CoP position in both the AP and gain setting on changes in CoP was increased to 0.5 (with ML ML directions. To maintain the leaning postures at the end the tuned gain and SPW settings for CoP shifts in the AP of the FO and FR paths, adjustments in the CoP position direction), no undesirable changes in posture were observed. were mainly elicited in the ML direction. To maintain the Therefore, the PFC gain for ML direction was set at 0.5 for leaning posture at the end of the FL path, adjustments in all repetitions. the CoP position were elicited in both the AP and ML direc- tions. The mean 95% PEA for leaning movements in the FO, 3.2. Controller Actions. The mean changes in CoP trajectories FR, and FL directions were 951 0 ± 341 1mm , 1095 9± and stimulation pulse widths for leaning postures in the FO 2 2 251 2mm , and 1364 5 ± 688 2mm , respectively. The 95% direction are represented in Figure 5. After about 1 second, PEA was the greatest for leaning movements in the FL direc- posture began to change from the nominal starting position tion, suggesting that the overall CoP position deviated from and arrived at the end of the path after approximately 2 sec- the moving circle when leaning toward the end of the path onds. The primary muscles activated during this movement and returning to NO, as illustrated in Figure 7(a). The predic- were the bilateral GS, as the greatest change in the CoP posi- tion ellipses for one repetition of leaning movements in each tion occurred in the AP direction. The leaning posture was direction are displayed in Figure 7(b). The 95% PEA for the maintained for about 3 seconds, during which, fluctuations leaning movement in the FO direction was 1276.5 mm , in the ML component of CoP were elicited to ensure posture 2 2 while the 95% PEA was 1141.8 mm and 1645.2 mm for was maintained. These postural adjustments resulted in acti- the leaning movement in FR and FL directions, respectively. vation to the left GMED and right PA at the beginning and The mean width of the prediction ellipses (Figure 8) for end of the dwell period. The left PA and right GMED were leaning movements in the FO, FR, and FL directions were activated in the middle of the dwell period. This suggests that 13 9±3 8mm, 16 1±4 3mm, and 17 7±8 6mm, respec- to maintain the FO leaning posture for this subject, adjust- tively. The FL direction has the largest ellipse width and ments towards the right were required at the beginning and greatest PEA, further suggesting that greater CoP deviations end of the dwell period. To maintain the FO-leaning posture from the moving circle occurred in that direction. during the middle of the dwell period, adjustments towards the left were required. During the dwell period, the largest changes in activations were to the bilateral GS, which each 4. Discussion reached a maximum of 16 μs. After the 3-second dwell The aim of this study was to implement the PFC in the labo- period, posture was adjusted back towards the nominal start- ratory setting and conduct an experimental feasibility test ing position. Activation to the bilateral GS decreased as the with a standing NP user. This is the first study to our knowl- nominal posture was attained. Activations to the bilateral edge to investigate the modulation of LE stimulation in a PAs and GMEDs were minimal during this portion of the standing NP user as posture is adjusted away from erect leaning movement. stance via a feedback controller. In this study, the feedback 3.3. Standing Performance. Standing duration was an average signal was the CoP position, which was readily obtained from of 1 minute and 55 seconds (±6 seconds) per trial. The mean force plates. As the subject leaned away from an erect stance, 8 Applied Bionics and Biomechanics Mean changes in CoP position (forward leaning postures) -20 0 1234567 -10 01234 567 Mean cycle time (s) (a) Mean changes in stimulation pulse widths for forward leaning postures 20 20 0 0 10 10 5 5 0 0 -5 -5 0 12 34 56 7 5 15 0 5 -5 -5 0 1234567 0123456 7 Mean cycle time (s) (b) Figure 5: Mean changes across the five trials in (a) the overall CoP position and (b) muscle stimulation pulse widths as posture was shifted in the forward direction. In (a), the mean CoP profiles are presented for the anterior-posterior (AP) direction and the medial-lateral (ML) direction. In (b), the changes in stimulation PWs are presented for the following muscles: LGS (left gastrocnemius), RGS (right gastrocnemius), LTA (left tibialis anterior), RTA (right tibialis anterior), LGMED (left gluteus medius), RGMED (right gluteus medius), LPA (left posterior adductor), and RPA (right posterior adductor). In all plots of the repetitions in the forward direction, the mean profiles are indicated with bold solid lines and (±1) standard deviation is indicated with dashed lines. the PFC modulated stimulation proportionally according to according to changes in the CoP position. While the move- the desire to effect postural change during the tracking tasks. ment strategies may vary slightly from trial to trial and con- Compared to the maximum resultant UE force exerted dition to condition, the overall strategy implemented was while the subject stood in the NO position during the static consistent over all the trials. Any variations in UE muscle trial (6.75% BW), large percent differences in mean maxi- activation from condition to condition would minimally mum resultant UE effort exerted were observed for all the affect the movement of the CoP and average out over the leaning directions (%difference ≥ 116 3%). In the simulation repeated trials. While the movement strategies may also vary study [10], the PFC reduced UE effort by an average of 50%, from subject to subject, the user acted at his own control, and compared with UE effort alone. UE contribution to leaning it is unlikely that voluntary UE muscle activation patterns postures was modeled as simple impedance forces defined would change significantly. No visual differences in the strat- as linear functions of the shoulder position. In these experi- egy implemented to adjust posture across trials were ments, the subject’s specific volitional strategy to use the observed; thus, changes in UE muscle activation were not UEs to elicit changes in posture was not controlled. The anticipated. The findings in this experimental study indicate PFC modulated activation of the paralyzed LE muscles only that there are greater demands placed on the UEs while LGS (s) LPA LGMED CoP ML (mm) CoP AP (mm) RGS RGMED RPA Applied Bionics and Biomechanics 9 Mean peak UE force extended during CoP tracking tasks right, left, backward, backward-right, and backward-left as an outcome measure). The average CoP was 20,181 8± area 2 2 4527 8mm compared to 19,332 4 ± 3557 1mm in their able-bodied subject group. We hypothesize that the PEAs 25 observed in our study are less than those reported in [20] due to several differences in the experimental design and sub- ject population. First, our experimental design required the subject to stand within an instrumented walker, which lim- ited how far CoP could be adjusted in each direction, while subjects in the Lemay experiments stood with their hands at their sides without the constraint of an enclosure. Second, NO FO FR FL our experimental design also required that our subject Leaning posture direction adjusts the CoP by tracking a moving circle at a fixed velocity. Lemay’s subjects had 15 seconds to complete the leaning Figure 6: Mean maximum resultant UE force during leaning movements in the forward (FO), forward-right (FR), and forward- movement at a self-selected speed. Third, our subject was left (FL) directions. As a reference, the maximum resultant UE nonambulatory, had no control of his ankle plantar/dorsi- force exerted while the subject stood in the nominal (NO) starting flexors without stimulation, and required a support device position during a static trial is also displayed. Error bars are to stand. The subjects with SCI in the Lemay study were com- included to indicate ±1 standard deviation of measurements munity ambulators who could stand for 5 minutes without a across twelve repetitions. support device (AIS D), and many had near normal walking ability (1.02 m/s) [20]. Thus, they may not be representative changing posture, suggesting the impedance model may be a of individuals with incomplete SCI. Lemay et al. [20] further highly simplified representation of UE contribution to lean- state this as a possible reason that there was no statistical dif- ing movements. Thus, future work should explore more ference found in CoP between the SCI and able-bodied accurate representations of the interaction forces between area the UEs and the support device during leaning movements. groups. Future work will repeat these experiments with addi- Another potentially confounding assumption in the model tional subjects to determine PEAs that are more representa- was that the feet were fixed to the ground and not allowed tive of nonambulatory individuals with incomplete SCI. to rotate. Thus, the simulation outcomes could freely apply The CoP position feedback, as measured with force high-activation levels to the ankle plantar flexors without plates, was a practical signal for laboratory-based exploratory causing the model to fall over. These might have been the experiments with the PFC. However, the long-term goal is to reasons for a high reduction in the UE forces in simulation, deploy the controller for home use. Force plates limit con- which was not practicable in real life because of the troller deployment to the laboratory setting, but advances heel-lifting effect observed for the current study participant. in sensor technology enable the accurate capture of body PEA and ellipse width were computed to determine motion outside of a controlled laboratory environment. CoP-tracking deviations. Across all leaning directions, the Insole-pressure measurement devices are an appealing PEA increased as deviations in CoP tracking occurred option for the measurement of CoP, given that the position (Figure 7). The ellipse width provided an additional measure of the feet on the floor relative to each other are specified. of CoP tracking, as it described how far the overall CoP devi- Each time the user stands, it is likely that the location of the ated from the prescribed straight-line path (Figure 8). The feet will differ slightly. This is not a major issue in the labora- mean peak resultant UE effort, PEA, and ellipse width were tory, where the feet can be moved to fixed targets before each all greatest during leaning movements in the FL direction, experiment. However, for implementation in the uncon- but were relatively similar for postural changes in the FO trolled environments of the home and community, addi- and FR directions. These findings may suggest that as the tional sensors would need to be added to determine the leaning posture deviated from the prescribed path, more distances between the feet and their orientation before com- UE effort may have been required to readjust posture towards puting the CoP position. The CoM position is a global vari- the path. The PFC continually updated stimulation to the LEs able that can be implemented to detect the position of the as the changes in posture were elicited. The differences in the body each time the user stands as well as to track the dynamic findings for the FR and FL directions may be attributed to changes in posture as the user prepares for a functional task. differences in UE strength or the individual differences in Furthermore, the CoM position more accurately reflects the the stimulated responses of each muscle, among other issues system dynamics and can change without commensurate dis- particular to this subject. Future work will repeat similar placements of the CoP. The CoM position is therefore an experiments with additional subjects. ideal parameter for controlling the entire system, particularly Lemay et al. [20] conducted the comfortable multidirec- for faster movements or to recover from perturbations. tional limits of stability test with visual feedback to investi- Methods to estimate the CoM position from a network of gate dynamic postural stability in ambulatory individuals body-mounted inertial measurement units are underdevel- with SCI. They reported CoP ,defined by an ellipse fitting oped, and future work will verify such techniques and incor- area the linear distance between the initial and maximal positions porate them into home-going systems employing the of the CoP in each of the eight tested directions (FO, FR, FL, whole-body CoM position as the feedback signal. UE force (% BW) 10 Applied Bionics and Biomechanics 95% Prediction ellipse area for tracking tasks Posturogram of tracking tasks -10 −10 -20 −20 −50 −40 −30 −20 −10 0 10 20 30 40 50 -40 -20 0 20 40 60 CoP in ML direction (mm) CoP ML (mm) (a) (b) Figure 7: (a) Typical posturogram and (b) 95% prediction ellipses for CoP-tracking tasks in the forward and diagonal directions. The 95% 2 2 2 PEA for the leaning movements in the forward direction was 1276.5 mm , 1141.8 mm in the forward-right direction, and 1645.2 mm in the forward-left direction. Based on the orientation of the laboratory coordinate system, postural adjustments in the forward direction are indicated as CoP increasing from the nominal starting position. Postural shifts towards the left are indicated as CoP increasing from AP ML the nominal. Mean width of 95% prediction ellipses or coronal planes. Future work should explore and exploit the coupling between muscle actions and include cross terms 25 to represent the effects of the GS and TA on ML movement and PA and GMED on AP movement. This involves extend- ing the PFC to act in the generalized coronal plane and mod- ulating all muscles simultaneously irrespective of assumed movement direction (including the postural muscles for hip extension/flexion or trunk extension/lateral bending not adjusted in the current study) to generate the globally opti- mal patterns of stimulation to realize a movement. This study sought to determine the experimental feasibil- FO FR FL ity of the PFC, a muscle activation controller that modulated Leaning posture direction LE activation according to changes in the CoP position, in a recipient of an implanted standing NP. The PFC enabled Figure 8: Mean widths of 95% prediction ellipses for CoP-tracking the subject to assume leaning postures in the FO, FR, and tasks in the forward (FO), forward-right (FR), and forward-left (FL) FL directions, by modulating LE muscle activation according directions. The error bars are included to indicate ±1 standard to changes in the overall CoP position. More than twice the deviation of measurements across twelve repetitions. UE effort as a percentage of quiet standing were required to effect changes in CoP experimentally in this study as pre- dicted from the simulations presented in [10]. CoP-tracking A limitation of this study is the length of time the subject could stand during the experiments. Although the subject results indicate that all paths presented were successfully could stand quietly for 25 minutes at the time of testing, these tracked, suggesting that the PFC provided the subject with experiments were more demanding because they entailed more access to the workspace while standing. multiple repetitions of standing and adjusting posture in the different directions. To minimize fatigue induced by con- 5. Conclusions tinuous activation of the muscles, the number of repetitions collected was limited, so that the subject’s total standing time We have explored the experimental feasibility of the PFC, a did not exceed 10 minutes. This is consistent with elapsed CoP-position tracking muscle activation controller with a standing times with conventional FNS systems [2]. recipient of an implanted standing NP. This is the first study Another limitation to this study is the availability of mus- to our knowledge that investigates feedback control of stand- cles for control as well as the directions in which the recruited ing posture to enable user-selected leaning movements away muscles acted. The PFC, as implemented, assumed that the from erect stance in an individual with SCI. As the CoP posi- muscles acted independently and exclusively in the sagittal tion was adjusted to track the moving circle along the various Ellipse Width (mm) CoP in AP direction (mm) CoP AP (mm) Applied Bionics and Biomechanics 11 [8] H. Rouhani, M. Same, K. Masani, Y. Q. Li, and M. R. Popovic, paths, the PFC continually updated activation to the user’s “PID controller design for FES applied to ankle muscles in paralyzed LE musculature. Ellipse areas of the CoP traces neuroprosthesis for standing balance,” Frontiers in Neurosci- indicate that the PFC provided the user with greater access ence, vol. 11, p. 347, 2017. to the standing workspace. Future work will evaluate the con- [9] R. Nataraj, M. L. Audu, and R. J. Triolo, “Center of mass accel- troller with the whole-body CoM position as the feedback eration feedback control of standing balance by functional signal and account for cross-coupling resulting from the ana- neuromuscular stimulation against external postural perturba- tomical actions of the contracting muscles. This will require tions,” IEEE Transactions on Biomedical Engineering, vol. 60, the development and evaluation of a model that outputs no. 1, pp. 10–19, 2013. CoM from data captured from body-mounted sensors and [10] M. L. Audu, S. J. Gartman, R. Nataraj, and R. J. Triolo, “Postur- more advanced multidimensional control algorithms. e-dependent control of stimulation in standing neuroprosth- esis: simulation feasibility study,” Journal of Rehabilitation Data Availability Research and Development, vol. 51, no. 3, pp. 481–496, 2014. [11] S. J. Gartman, M. L. Audu, R. F. Kirsch, and R. J. Triolo, “Selec- The data used to support the findings of this study are tion of optimal muscle set for 16-channel standing neuro- available from the corresponding author upon request. prosthesis,” The Journal of Rehabilitation Research and Development, vol. 45, no. 7, pp. 1007–1018, 2008. Conflicts of Interest [12] J. J. Abbas and J. C. Gillette, “Using electrical stimulation to control standing posture,” IEEE Control Systems, vol. 21, The authors declare that there is no conflict of interest no. 4, pp. 80–90, 2001. regarding the publication of this paper. [13] N. Bhadra, K. L. Kilgore, and P. H. Peckham, “Implanted stim- ulators for restoration of function in spinal cord injury,” Med- ical Engineering & Physics, vol. 23, no. 1, pp. 19–28, 2001. Acknowledgments [14] M. A. Schiefer, K. H. Polasek, R. J. Triolo, G. C. J. Pinault, and This material is also the result of work supported with D. J. Tyler, “Selective stimulation of the human femoral nerve resources and the use of facilities at the Louis Stokes Cleveland with a flat interface nerve electrode,” Journal of Neural Engi- Veterans Affairs Medical Center in Cleveland, OH. Funding neering, vol. 7, no. 2, article 26006, p. 026006, 2010. for this work was provided by the National Institute of Neuro- [15] R. J. Triolo, C. Bieri, J. Uhlir, R. Kobetic, A. Scheiner, and E. B. logical Disorders and Stroke (Grant No: R01NS040547) and Marsolais, “Implanted functional neuromuscular stimulation the Craig H. Neilsen Foundation (Grant No: 459308). systems for individuals with cervical spinal cord injuries: clin- ical case reports,” Archives of Physical Medicine and Rehabili- tation, vol. 77, no. 11, pp. 1119–1128, 1996. References [16] B. Smith, Zhengnian Tang, M. W. Johnson et al., “An externally powered, multichannel, implantable stimulator- [1] G. P. Forrest, T. C. Smith, R. J. Triolo et al., “Use of the case telemeter for control of paralyzed muscle,” IEEE Transactions Western reserve/veterans administration neuroprosthesis for on Biomedical Engineering, vol. 45, no. 4, pp. 463–475, 1998. exercise, standing and transfers by a paraplegic subject,” Dis- ability and Rehabilitation: Assistive Technology, vol. 7, no. 4, [17] D. A. Winter, F. Prince, J. S. Frank, C. Powell, and K. F. Zabjek, pp. 340–344, 2012. “Unified theory regarding A/P and M/L balance in quiet stance,” Journal of Neurophysiology, vol. 75, no. 6, pp. 2334– [2] R. J. Triolo, S. N. Bailey, M. E. Miller et al., “Longitudinal per- 2343, 1996. formance of a surgically implanted neuroprosthesis for lower-extremity exercise, standing, and transfers after spinal [18] S. J. Gartman, M. L. Audu, R. F. Kirsch, and R. J. Triolo, “Selec- cord injury,” Archives of Physical Medicine and Rehabilitation, tion of optimal muscle set for 16-channel standing functional vol. 93, no. 5, pp. 896–904, 2012. electrical stimulation system,” Presented at the 12th Ann. [3] J. J. Abbas and H. J. Chizeck, “Feedback control of coronal Conf. Int. FES Soc, Philadelphia, PA, USA, 2007. plane hip angle in paraplegic subjects using functional neuro- [19] P. Schubert and M. Kirchner, “Ellipse area calculations and muscular stimulation,” IEEE Transactions on Biomedical Engi- their applicability in posturography,” Gait and Posture, neering, vol. 38, no. 7, pp. 687–698, 1991. vol. 39, no. 1, pp. 518–522, 2014. [4] M. Moynahan and H. J. Chizeck, “Characterization of paraple- [20] J.-F. Lemay, D. H. Gagnon, S. Nadeau, M. Grangeon, gic disturbance response during FNS standing,” IEEE Transac- C. Gauthier, and C. Duclos, “Center-of-pressure total trajec- tions on Rehabilitation Engineering, vol. 1, no. 1, pp. 43–48, tory length is a complementary measure to maximum excur- sion to better differentiate multidirectional standing limits of [5] R. J. Jaeger, “Design and simulation of closed-loop electrical stability between individuals with incomplete spinal cord stimulation orthoses for restoration of quiet standing in para- injury and able-bodied individuals,” Journal of NeuroEngineer- plegia,” Journal of Biomechanics, vol. 19, no. 10, pp. 825–835, ing and Rehabilitation, vol. 11, no. 1, p. 8, 2014. [6] K. J. Hunt, H. Gollee, and R. P. Jaime, “Control of paraplegic ankle joint stiffness using FES while standing,” Medical Engi- neering & Physics, vol. 23, no. 8, pp. 541–555, 2001. [7] Z. Matjacic and T. Bajd, “Arm-free paraplegic standing- part II: experimental results,” IEEE Transactions on Rehabilitation Engineering, vol. 6, no. 2, pp. 139–150, 1998. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Applied Bionics and Biomechanics Hindawi Publishing Corporation

Experimental Implementation of Automatic Control of Posture-Dependent Stimulation in an Implanted Standing Neuroprosthesis

Loading next page...
 
/lp/hindawi-publishing-corporation/experimental-implementation-of-automatic-control-of-posture-dependent-L0IvHdoECH
Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2019 Brooke M. Odle et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
1176-2322
eISSN
1754-2103
DOI
10.1155/2019/2639271
Publisher site
See Article on Publisher Site

Abstract

Hindawi Applied Bionics and Biomechanics Volume 2019, Article ID 2639271, 11 pages https://doi.org/10.1155/2019/2639271 Research Article Experimental Implementation of Automatic Control of Posture-Dependent Stimulation in an Implanted Standing Neuroprosthesis 1,2 2 1,2 1,2 Brooke M. Odle , Lisa M. Lombardo , Musa L. Audu , and Ronald J. Triolo Department of Biomedical Engineering, Case Western Reserve University, Cleveland 44106, USA Motion Study Laboratory, Louis Stokes Cleveland Veterans Affairs Medical Center, Cleveland 44106, USA Correspondence should be addressed to Brooke M. Odle; brooke.odle@case.edu Received 23 May 2018; Accepted 13 January 2019; Published 14 March 2019 Academic Editor: Le Ping Li Copyright © 2019 Brooke M. Odle et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Knowledge of the upper extremity (UE) effort exerted under real-world conditions is important for understanding how persons with motor or sensory disorders perform the postural shifts necessary to complete many activities of daily living while standing. To this end, a feedback controller, named the “Posture Follower Controller”, was developed to aid in task-dependent posture shifting by individuals with spinal cord injury standing with functional neuromuscular stimulation. In this experimental feasibility study, the controller modulated activation to the paralyzed lower extremity muscles as a function of the position of overall center of pressure (CoP), which was prescribed to move in a straight line in forward and diagonal directions. Posture-dependent control of stimulation enabled leaning movements that translated the CoP up to 48 mm away from the nominal position during quiet standing. The mean 95% prediction ellipse area, a measure of the CoP dispersion in the forward, 2 2 2 forward-right, and forward-left directions, was 951 0 ± 341 1mm , 1095 9 ± 251 2mm , and 1364 5 ± 688 2mm , respectively. The average width of the prediction ellipses across the three directions was 15.1 mm, indicating that the CoP deviated from the prescribed path as task-dependent postures were assumed. The average maximal UE effort required to adjust posture across all leaning directions was 24.1% body weight, which is only slightly more than twice of what is required to maintain balance in an erect standing posture. These preliminary findings suggest that stimulation can be modulated to effectively assume user-specified, task-dependent leaning postures characterized by the CoP shifts that deviate away from the nominal position and which require moderate UE effort to execute. 1. Introduction device, such as a walker or a countertop. To address this lim- itation, previous groups explored closed-loop feedback con- trol systems for standing with stimulation employed at Spinal cord injury (SCI) often results in partial or total individual joints [3–8] as well as a stimulation controller paralysis of the trunk and lower extremity (LE) muscles. Implanted neuroprostheses (NPs) utilizing functional neuro- based on comprehensive or global joint feedback combined with center of mass (CoM) acceleration that rejected destabi- muscular stimulation (FNS) can restore basic standing func- lizing perturbations and reduced the UE effort to maintain tion in individuals with SCI, providing them with the standing balance [9]. However, these advanced control sys- independence to accomplish several activities of daily living tems have been designed to maintain only a single upright [1, 2]. Standing NPs supply constant preprogrammed open- loop stimulation to the trunk, hip, and knee extensors to setpoint in the nominal standing position. Users are only able to stand optimally and resist potentially destabilizing pertur- maintain a single, upright stance. Thus, to maintain balance bations in one erect, neutral posture rather than at forward- in the presence of postural perturbations, NP users rely on or side-leaning postures best suited for specific functional voluntary upper extremity (UE) effort exerted on a support 2 Applied Bionics and Biomechanics The aims of this work are (1) to explore the effectiveness tasks [10, 11]. To adjust posture away from the erect stance with existing systems that deliver preprogrammed patterns of a PFC to enable NP users to lean away from the erect pos- of stimulation, users must exert voluntary UE effort to push ture and (2) to examine the contribution of UE effort to lean- ing postures in real subjects with SCI using a NP for standing. or pull against the support device. At these new positions, the patterns of stimulation tuned for erect standing become In this exploratory study, standing performance was deter- suboptimal and may over or understimulate the muscles mined by the following metrics: (1) maximum resultant UE required to maintain the new task-dependent postures. effort contributed to leaning movements in the AP and ML One benefit to the ability to adjust posture from the erect directions as LE muscle activation is modulated and (2) CoP-tracking deviations from a prescribed straight-line path. stance is the ability to prepare for a functional task, such as reaching and manipulating objects on shelves. This also gives users the ability to reach the full extent of their standing 2. Materials and Methods workspace, thereby providing them with greater indepen- dence and access to objects in the environment. Another ben- 2.1. Subject and Standing Neuroprosthesis System. A 27-year- efit is adjusting posture laterally, to rest muscles on one side old male with motor incomplete C5 tetraplegia (AIS C) par- of the body as a means of mitigating fatigue and prolonging ticipated in the experiments. He was approximately 185.4 cm overall standing times. These benefits are further supported tall and weighed 58.5 kilograms when the experiments were by the work of Abbas and Gillette [12], which suggested that conducted. He received a 16-channel implanted LE NP one shifting posture in the anterior-posterior (AP) and medial- year prior to data collection and was a regular user for recon- lateral (ML) directions would be critical in enabling standing ditioning exercise and standing. At the time of the study, he NP users to accomplish various functional tasks. could stand quietly in the neutral position for 25 minutes To achieve these changes in posture, the location of the with 93% body weight (BW) supported by his legs, utilizing projection of the total body CoM on the base of support will UE effort only for a light touch on the walker to maintain bal- have to change smoothly and continuously. Thus, control ance. Prior to participating in the experiments, the subject systems that can automatically maintain standing balance signed informed consent forms approved by the Institutional should include CoM position feedback and modulate Review Board of the Louis Stokes Cleveland Veterans Affairs stimulation to the LE muscles as posture is adjusted away Medical Center. from the erect stance. One such system (the Posture Follower The standing NP consisted of one surgically implanted Conftroller or PFC) was developed and tested in simulation 16-channel stimulator telemeter [13]. The implanted system [10]. In that simulation feasibility study, the model exerted targeted the following muscle groups: hip extensors (Gluteus voluntary UE effort to manually adjust the CoM location Maximus (GMX), Hamstring (HM), and posterior portion of from an initial neutral setpoint towards the new desired for- the Adductor Magnus (PA)), hip abductors (left Gluteus ward- or side-leaning posture, while the controller continu- Medius (GMED)), trunk extensors (Lumbar Erector Spinae ally updated the neural stimulation to maintain activations (ES)), and the trunk lateral benders (Quadratus Lumborum that were optimal for each change in position. Controller per- (QL)). A selective, multicontact, flat interface nerve cuff elec- formance was measured with respect to (1) the UE effort, trode [14] was implanted on the proximal femoral nerve near defined as the UE forces exerted as the simulated user chan- the inguinal ligament, to activate the three uniarticular vasti ged postures and (2) the ability to track a moving object with muscles of the quadriceps (QD) while avoiding recruitment the CoM as the object moved in the forward, diagonal, and of the sartorius and the biarticulated rectus femoris, which lateral directions. The PFC reduced UE effort by an average induce hip flexion that compromises erect neutral standing. of 50%, compared to using the UEs alone. In general, CoM All other muscles were activated with surgically implanted tracking with the PFC and UE alone was similar, except for intramuscular electrodes [15]. Pulse amplitudes (0.8, 2.1, one instance of an overshoot when moving in the left- 18, or 20 mA) were set on a channel-by-channel basis while diagonal direction without the active controller. These pulse duration (0-250 μs) and frequency (0-20 Hz) were encouraging simulation findings were encouraging and sup- modulated independently on a pulse-by-pulse basis on each port the development of posture-dependent control of stim- channel to achieve the desired motion. A rechargeable ulation for a standing NP. wearable external control unit (ECU) [16] delivered power The next step in the development and deployment of and command signals to the implanted pulse generator via posture-dependent control systems is to implement the a close-coupled inductive link maintained by a transmitting PFC in the laboratory environment with a standing NP user. coil taped to the skin over the stimulator. The ECU coordi- The simulation results in [10] used an ideal UE controller nated the delivery of temporal patterns of stimulation that modulated UE forces as a function of shoulder displace- through all 16 channels simultaneously. ment and velocity away from the erect posture position. The To target the right GMED for postural control in the ML actual UE forces exerted by a standing NP user are generally direction and bilateral tibialis anterior (TA) and gastrocne- obtained from measuring devices in the laboratory and could mius (GS) for postural control in the AP direction, the vary considerably between individuals. Knowledge of the implanted stimulation system was supplemented with self- actual UE effort required by standing NP users to change adhesive surface electrodes. These muscles were recruited posture will greatly help in the design of control systems for because they were not available in the subject’s implanted ensuring balance during performance of important activities system. Surface stimulation was delivered at a constant fre- of daily living. quency of 20 Hz, variable pulse width up to 250 μs dependent Applied Bionics and Biomechanics 3 on controller output, and a fixed pulse amplitude (100 mA where CoP and CoP are the AP and ML components of AP ML for the right GMED and bilateral GS, and 30 mA for the bilat- the CoP. CoP , CoP , CoP , and CoP are the x and y Lx Ly Rx Ry eral TA). components of the CoP, and F and F are the vertical reac- L R Real-time control of stimulation was implemented with a tion forces under the left and right feet, respectively. custom software developed in MATLAB/Simulink R7.9 and As the subject leans away from the erect stance and the the xPC Target toolbox (MathWorks Inc., Natick, MA). A CoP moves away from the nominal (erect) position, the Windows (Microsoft Inc., Redmond, WA) host computer resulting changes in AP and ML components are tracked was utilized to build customized applications, while a dedi- by the PFC via a simple proportional control law. Assuming cated (target) computer with the Pentium Dual-Core 3 GHz a linear relationship between changes in the CoP and muscle microprocessor (Intel Inc., Santa Clara, CA) with 2 GB of activation, the changes in activation to be applied to the LE muscles are computed according to (equation 3): RAM was responsible for running the applications in real time. The host and target computers communicated via the TCP/IP protocol. Data were acquired using a NI PCI-6071E SPW‐BPW ∗ ΔCoP Muscle activation = BPW + Gain ∗ , board (National Instruments Inc., Austin, TX). For the Max CoP iD experiments described, all real-time controller and stimula- tion parameters were sampled at 40 Hz. The stimulus values for erect standing were determined by clinical observation whereby the subject exhibited ample knee, hip, and trunk where BPW is the baseline pulse width value, Gain is the extension to achieve an erect posture without discomfort. proportional gain applied to the AP and ML components of Baseline standing stimulation values are listed in Table 1. the directions of the overall CoP, SPW is the muscle satura- tion pulse width value (defined as the maximum above which 2.2. Posture Follower Controller Design. In this study, the no additional force is generated), CoP is the subject’s instan- overall center of pressure (CoP) position (a function of the taneous CoP position, and Max CoP is the maximum iD location of the vertical ground reaction force vector) was excursion of the path of the object being tracked by the sub- used as the feedback signal for the PFC. A more suitable feed- ject’s CoP relative to the nominal position. In this equation, i back signal would be the orthogonal projection of the is a placeholder for either the AP or ML component of CoP. whole-body CoM or center of gravity (CoG). However, there Thus, ΔCoP is the difference between the current CoP posi- are challenges in implementing the CoM position (a function tion and the nominal CoP position in the AP or ML direc- of the location of the total body mass, as the feedback signal). tion. The gain setting was limited to values between 0 and Currently, there is no means for the quantity to be computed 1, to prevent the muscle activation from exceeding SPW. or estimated from body-mounted sensors in real time rapidly Assuming that posture is adjusted in a slow and and accurately enough to use as a stimulus control signal quasi-static manner, the PFC targets muscles to provide sup- with a paralyzed user. The CoM, CoG, and CoP are equiva- port, supplying stimulation that is optimal, as determined in lent during static conditions. Thus, the overall CoP position [10], for the static position at any time. Previous simulation was used as a surrogate because it can be readily obtained studies [10, 18] determined that the bilateral GS, GMED, from two force plates (AMTI, Watertown, MA) in the labora- and PA provide support as the postural shifts are elicited tory, making it a more practical control signal for this explor- away from the erect stance. Based on these findings, the atory study. The laboratory-based PFC took the form of a PFC only modulates activation to those muscles. The GS proportional feedback controller, so it tracks voluntary are ankle plantar flexors and were modulated as posture changes in posture by mapping changes in CoP to changes was adjusted forward and backward in the sagittal plane. in LE muscle activations (Figure 1). The TAs are ankle dorsiflexors and were activated at a fixed The user stood at an erect, nominal stance with baseline pulse width to cocontract with the GS, to increase ankle joint (open-loop) stimulation. In this stance, the user stood upright stiffness or, as we observed in the current subject, to help mit- with the feet approximately under the shoulders and each on a igate any spasms that might be triggered as posture is separate force plate. The erect, nominal stance is biomechan- adjusted. The GMED and PA are hip ab/adductors, respec- ically defined as the standing posture in which the head, tively, and were utilized to effect postural shifts in the ML trunk, pelvis, and LEs are aligned as close to vertical as possi- direction. When posture was adjusted leftward, the right ble in sagittal and coronal planes with minimal to no axial GMED and left PA were targeted for activation to support rotation in the coronal plane. The components of the overall the body against the pull of gravity. Conversely, when pos- CoP position in the AP and ML direction were computed ture was adjusted rightward, the left GMED and right PA using (equation (1)) and (equation (2)), respectively [17]. were targeted for activation. The PFC did not modulate mus- cle activation to the extensor muscles (right HM, right ES, F F L R CoP = CoP ∗ + CoP ∗ , 1 and bilateral GMX and QD) which had to be maximally stim- AP Lx Rx F +F F +F L R L R ulated to maintain an erect standing posture. 2.3. Visual Feedback. To assess the standing performance F F L R CoP =CoP ∗ + CoP ∗ , ML Ly Ry with respect to maintaining and tracking posture according F +F F +F L R L R to a prescribed path, visual feedback of the overall CoP and specified paths were presented on a computer monitor in real 4 Applied Bionics and Biomechanics Table 1: Muscle pulse amplitudes and pulse widths for baseline standing. Muscles that were always recruited for baseline standing are indicated with a “P”, while those recruited by the controller are indicated with a “C.” Muscles that were supplemented with surface electrodes are indicated with a “ ”. Baseline standing Muscle Function Pulse amplitude (mA) Threshold PW (μs) Saturation PW (μs) PW (μs) Right gluteus maximus (right GMX) P 20.0 248 2 250 Right hamstring (right HM) P 20.0 250 64 250 Left gluteus maximus (left GMX) P 20.0 145 5 150 Left gluteus medius (left GMED) C 20.0 61.5 13 110 Right quadratus lumborum (right QL) —— — — — Right erector spinae (right ES) P 2.1 90 10 125 Left quadratus lumborum (left QL) — 18.0 0 10 50 Left erector spinae (left ES) — 18.0 0 35 70 Right quadriceps 1 (right QD 1) P 0.8 90 30 90 Right quadriceps 2 (right QD 2) P 0.8 90 24 90 Left quadriceps 1 (left QD 1) P 0.8 250 64 250 Left quadriceps 2 (left QD 2) P 0.8 250 48 250 Right quadriceps 3 (right QD 3) P 0.8 100 32 100 Right posterior adductor (right PA) C 20.0 86 2 170 Left quadriceps 3 (left QD 3) P 0.8 250 72 250 Left posterior adductor (left PA) C 20.0 128.5 7 250 Right gluteus medius (right GMED) 100.0 0 80 250 Right tibialis anterior (right TA) 30.0 0 80 100 Right gastrocnemius (right GS) 100.0 0 20 65 Left tibialis anterior (left TA) C 30.0 0 70 90 Left gastrocnemius (left GS) 100.0 0 30 70 CoP to muscle User’s Muscle activation Erect stance lower activation conversion extremities (Nominal CoP position) − (posture follower controller) Actual CoP position (CoP AP, CoP ML) Figure 1: Control setup. The user stands erect on force plates, which measure the center of pressure (CoP) position in the anterior-posterior (AP) and medial-lateral (ML) directions. The user leans away from the erect stance, adjusting the overall CoP position towards the ends of the paths in the forward and diagonal directions. The force plates continuously track the resulting changes in the CoP position and the posture follower controller converts the changes in the CoP position to muscle activation, which is applied to the lower extremities. time (Figure 2). To ensure that posture was adjusted at a con- 40 mm and 48 mm from the nominal starting position in the forward and diagonal directions, respectively. sistent rate, the subject adjusted his CoP to track a circle moving along a straight line on the computer screen. The moving circle traveled at a speed of 20 mm/s to the end of 2.4. Posture Follower Controller Tuning. The proportional the selected path and returned to the nominal position along gain settings in (equation 3) were tuned by hand over several the same path. This speed was selected because it was the experimental sessions. To determine the optimal settings, the subject stood erect on the force plates with visual feedback maximum speed that enabled the subject to adjust CoP in a continuous manner. The length of each of the paths was and adjusted posture to place his CoP at the location of the based on the subject’s comfort while leaning forward and endpoints of the paths in the forward and diagonal direc- diagonally. The endpoints of each path were positioned tions. If the gain settings were too high, activation to the Applied Bionics and Biomechanics 5 End of forward (FO) path End of forward-right (FR) path (end of current specified path) Current specified path End of forward- le (FL) path Moving circle (tracking signal) Subject’s CoP Nominal (NO) starting position Figure 2: Diagram of the visual feedback display. The subject stood erect at the nominal (NO) starting position and adjusted the overall center of pressure (CoP) position to track the moving circle to the end of the paths defined by the yellow circles. The subject tracked the moving circle along the same path to return to the NO position. Prior to conducting the experiments, the speed of the moving circle and the locations of the endpoints of the paths were tuned to ensure that the subject adjusted posture at a comfortable rate and within reasonable limits of his standing balance. During the experiments, the subject adjusted the overall CoP position (green) to track the moving circle (blue) in the forward (FO), forward-right (FR), and forward-left (FL) directions. As an additional visual cue, the currently specified path is defined by changing the color of its endpoint from yellow to red. In this image, the currently specified path is the one from NO to FR. targeted muscles may require the subject to exert higher UE 2.6. Data Analysis. A repetition is distinguished by leaning forces to resist the muscle actions. Conversely, if the gain set- movement onset and offset (Figure 4). Movement onset was tings were too low, the controller would only nominally mod- defined as the initial time point where the moving circle departed from the nominal starting position. Upon reaching ulate activation to the LEs and diminish its potential impact on the posture. The gain values that enabled the subject to the end of the path, the moving circle dwelled there for 3 sec- comfortably adjust posture to and from the ends of all the onds before returning to the nominal starting position. paths were implemented in all the subsequent repetitions Movement offset was defined as the time point where the with the moving circle. moving circle first acquired the nominal starting position on its return. 2.5. Data Capture. The setup for testing the effects of the PFC There were twelve repetitions in which the subject con- is depicted in Figure 3. After activating baseline stimulation sistently maintained the starting position before movement to transition from sitting to upright standing, the subject onset, tracked the moving circle the entire distance to the donned a suspension harness (McMaster-Carr Inc., Elm- end, and maintained the same nominal starting position hurst, IL) attached to a lanyard (Guardian Fall Protection, after movement offset. Those repetitions were selected for Kent, WA) connected to a hook bolted into the laboratory analysis, and the overall CoP profiles were computed. To ceiling decking for safety. The subject stood with his hands obtain the changes in the CoP relative to the value at the nominal position, the starting CoP position was subtracted on a custom-built adjustable-instrumented walker (80/20, Columbia City, IN), which was adjusted for his height and from the resulting trajectories. UE effort, defined as the comfort. The subject also stood with each foot placed on a maximum resultant UE forces exerted with the PFC, was separate force platform to compute the overall CoP posi- compared to the values exerted during the erect stance tion in the AP and ML directions (equations 1 and 2). (equation 4). Upon settling into a comfortable erect standing position, the locations of the subject’s feet were marked with a tape UE − UE NO D Percent difference = to ensure the same foot placement throughout the experi- ∗ 100, 4 UE NO mental session. A static trial was collected to obtain the UE forces exerted on the walker at the nominal erect standing posture. where UE is the maximum UE force exerted at the nom- NO After instruction and sufficient practice to obviate learning inal position, and UE is the mean maximal UE effort effects, the subject adjusted posture by exerting volitional exerted while changing posture along the three directions UE effort on the instrumented walker to ensure his overall (FO, FR, and FL). Given the quasi-static nature of the CoP tracked the moving circle as it moved in the forward tracking tasks performed in this feasibility study, the 95% (FO) and diagonal directions (forward-right, FR; forwar- prediction ellipse area (PEA) was computed to describe d-left, FL). Five trials were collected, with two repetitions the dispersion of the CoP position in the directions investi- for tracking the circle to the ends of each of the three paths gated. The 95% prediction ellipse represents a region that and returning to the nominal erect position completed per contains the center of the points of the postural sway with trial. The sequence of directions was randomized to avoid 95% probability. Schubert and Kirchner [19] recommended systematic error. the PEA as a standard method of measuring posturographical 6 Applied Bionics and Biomechanics VICON cameras Safety lanyard and harness Computer monitor (visual feedback display) Reflective markers Instrumented walker Force plates Figure 3: Set-up for experimental evaluation of the controller. The subject stands erect on force plates, while holding onto an instrumented walker and adjusting the overall center of pressure (CoP) position towards the end of paths in the forward and diagonal directions. The subject was provided with visual feedback while adjusting the overall CoP position. Reflective markers were mounted on the subject to track his joint positions as he adjusted posture. CoP profiles during tracking task Dwell Dwell Movement Movement onset onset Movement Movement offset offset 105 110 115 120 125 130 Dwell Dwell Movement Movement Movement Movement onset onset offset offset 105 110 115 120 125 130 Trial time (secs) Subject Moving target Figure 4: The sample CoP profiles of the subject (blue) and the moving circle (red) of the two consecutive leaning movements during the tracking task. The top panel displays the CoP profiles in the AP direction, while the bottom panel displays the CoP profiles in the ML direction. Each movement is considered a separate repetition, which consists of a movement onset, dwell period, and movement offset. Movement onsets are indicated as the time point in which the moving circle initiates movement from the nominal starting position to the end of the specified path. Upon reaching the end of the specified path, the moving circle dwells there for 3 seconds. When the dwell period ends, the moving circle returns to the nominal position and remains there until it initiates travel along the next path. The first time point at which the moving circle acquires the nominal starting position on the return is the movement offset. Based on the orientation of the laboratory coordinate system, postural adjustments in the forward direction are indicated as CoP increasing from the nominal AP starting position. The postural shifts towards the left are indicated as CoP increasing from the nominal. Thus, in both repetitions, the ML subject was tracking the moving circle in the forward-left direction. CoP ML (mm) CoP AP (mm) Applied Bionics and Biomechanics 7 maximum resultant UE forces exerted while changing pos- scatter data, instead of the confidence ellipse. PEA was computed as shown in (equation 5). ture were computed for each leaning direction and normal- ized as percentage of BW (Figure 6). The mean maximum resultant UE effort exerted while eliciting leaning movements PEA = π ∗ χ ∗ det S , 5 in the FO, FR, and FL directions were 22 5±0 9%BW, 14 6 ±4 1%BW, and 35 2± 1 3%BW, respectively. As a reference, where χ is the inverse of the chi-square cumulative distribu- the maximum resultant UE force (6.75% BW) exerted while tion function with 2 degrees of freedom at a fixed probability standing in the nominal (NO) starting position is also dis- level (P = 95%, χ ≈ 5 99146), det(S) is the determinant 20 95,2 played. Compared to the maximum resultant UE force of the Eigenvalues of the sample variance covariance matrix exerted at NO, the percent difference in the mean maximum of CoP and CoP . The PEA and the width of each predic- resultant UE effort exerted during leaning movements in AP ML tion ellipse, a measure of CoP deviation from specified path, the FO, FR, and FL directions was 233.3%, 116.3%, and quantified the CoP-tracking performance. 421.5%, respectively. CoP excursions in the AP and ML directions during the 3. Results CoP-tracking task are displayed in Figure 7(a) and are repre- sentative of one repetition of leaning movements in each 3.1. Controller Tuning. When movements were elicited in the direction. To adjust posture in the FO direction, changes in forward or diagonal directions at gain settings larger than 0.4 CoP position in the AP direction were mainly required, with of the changes in CoP , the modulated activation to the AP minimal changes to CoP in the ML direction (as also indi- bilateral GS resulted in raising the heels off the ground so cated in Figure 5(a)). To lean in the FR direction, posture the subject stood on his toes. This heel-raising effect was was adjusted about 10 mm more in the ML direction than diminished when the gain was set to values below 0.35 and the AP direction. While leaning in the FL direction, posture the SPWs of the right and left GS were reduced from 100 μs was adjusted about the same distance in the AP and ML to 65 μs and from 90 μsto 70 μs, respectively. The tuned directions. However, maintaining posture at the end of each SPW for the target muscles are listed in Table 1. When the path required adjusting CoP position in both the AP and gain setting on changes in CoP was increased to 0.5 (with ML ML directions. To maintain the leaning postures at the end the tuned gain and SPW settings for CoP shifts in the AP of the FO and FR paths, adjustments in the CoP position direction), no undesirable changes in posture were observed. were mainly elicited in the ML direction. To maintain the Therefore, the PFC gain for ML direction was set at 0.5 for leaning posture at the end of the FL path, adjustments in all repetitions. the CoP position were elicited in both the AP and ML direc- tions. The mean 95% PEA for leaning movements in the FO, 3.2. Controller Actions. The mean changes in CoP trajectories FR, and FL directions were 951 0 ± 341 1mm , 1095 9± and stimulation pulse widths for leaning postures in the FO 2 2 251 2mm , and 1364 5 ± 688 2mm , respectively. The 95% direction are represented in Figure 5. After about 1 second, PEA was the greatest for leaning movements in the FL direc- posture began to change from the nominal starting position tion, suggesting that the overall CoP position deviated from and arrived at the end of the path after approximately 2 sec- the moving circle when leaning toward the end of the path onds. The primary muscles activated during this movement and returning to NO, as illustrated in Figure 7(a). The predic- were the bilateral GS, as the greatest change in the CoP posi- tion ellipses for one repetition of leaning movements in each tion occurred in the AP direction. The leaning posture was direction are displayed in Figure 7(b). The 95% PEA for the maintained for about 3 seconds, during which, fluctuations leaning movement in the FO direction was 1276.5 mm , in the ML component of CoP were elicited to ensure posture 2 2 while the 95% PEA was 1141.8 mm and 1645.2 mm for was maintained. These postural adjustments resulted in acti- the leaning movement in FR and FL directions, respectively. vation to the left GMED and right PA at the beginning and The mean width of the prediction ellipses (Figure 8) for end of the dwell period. The left PA and right GMED were leaning movements in the FO, FR, and FL directions were activated in the middle of the dwell period. This suggests that 13 9±3 8mm, 16 1±4 3mm, and 17 7±8 6mm, respec- to maintain the FO leaning posture for this subject, adjust- tively. The FL direction has the largest ellipse width and ments towards the right were required at the beginning and greatest PEA, further suggesting that greater CoP deviations end of the dwell period. To maintain the FO-leaning posture from the moving circle occurred in that direction. during the middle of the dwell period, adjustments towards the left were required. During the dwell period, the largest changes in activations were to the bilateral GS, which each 4. Discussion reached a maximum of 16 μs. After the 3-second dwell The aim of this study was to implement the PFC in the labo- period, posture was adjusted back towards the nominal start- ratory setting and conduct an experimental feasibility test ing position. Activation to the bilateral GS decreased as the with a standing NP user. This is the first study to our knowl- nominal posture was attained. Activations to the bilateral edge to investigate the modulation of LE stimulation in a PAs and GMEDs were minimal during this portion of the standing NP user as posture is adjusted away from erect leaning movement. stance via a feedback controller. In this study, the feedback 3.3. Standing Performance. Standing duration was an average signal was the CoP position, which was readily obtained from of 1 minute and 55 seconds (±6 seconds) per trial. The mean force plates. As the subject leaned away from an erect stance, 8 Applied Bionics and Biomechanics Mean changes in CoP position (forward leaning postures) -20 0 1234567 -10 01234 567 Mean cycle time (s) (a) Mean changes in stimulation pulse widths for forward leaning postures 20 20 0 0 10 10 5 5 0 0 -5 -5 0 12 34 56 7 5 15 0 5 -5 -5 0 1234567 0123456 7 Mean cycle time (s) (b) Figure 5: Mean changes across the five trials in (a) the overall CoP position and (b) muscle stimulation pulse widths as posture was shifted in the forward direction. In (a), the mean CoP profiles are presented for the anterior-posterior (AP) direction and the medial-lateral (ML) direction. In (b), the changes in stimulation PWs are presented for the following muscles: LGS (left gastrocnemius), RGS (right gastrocnemius), LTA (left tibialis anterior), RTA (right tibialis anterior), LGMED (left gluteus medius), RGMED (right gluteus medius), LPA (left posterior adductor), and RPA (right posterior adductor). In all plots of the repetitions in the forward direction, the mean profiles are indicated with bold solid lines and (±1) standard deviation is indicated with dashed lines. the PFC modulated stimulation proportionally according to according to changes in the CoP position. While the move- the desire to effect postural change during the tracking tasks. ment strategies may vary slightly from trial to trial and con- Compared to the maximum resultant UE force exerted dition to condition, the overall strategy implemented was while the subject stood in the NO position during the static consistent over all the trials. Any variations in UE muscle trial (6.75% BW), large percent differences in mean maxi- activation from condition to condition would minimally mum resultant UE effort exerted were observed for all the affect the movement of the CoP and average out over the leaning directions (%difference ≥ 116 3%). In the simulation repeated trials. While the movement strategies may also vary study [10], the PFC reduced UE effort by an average of 50%, from subject to subject, the user acted at his own control, and compared with UE effort alone. UE contribution to leaning it is unlikely that voluntary UE muscle activation patterns postures was modeled as simple impedance forces defined would change significantly. No visual differences in the strat- as linear functions of the shoulder position. In these experi- egy implemented to adjust posture across trials were ments, the subject’s specific volitional strategy to use the observed; thus, changes in UE muscle activation were not UEs to elicit changes in posture was not controlled. The anticipated. The findings in this experimental study indicate PFC modulated activation of the paralyzed LE muscles only that there are greater demands placed on the UEs while LGS (s) LPA LGMED CoP ML (mm) CoP AP (mm) RGS RGMED RPA Applied Bionics and Biomechanics 9 Mean peak UE force extended during CoP tracking tasks right, left, backward, backward-right, and backward-left as an outcome measure). The average CoP was 20,181 8± area 2 2 4527 8mm compared to 19,332 4 ± 3557 1mm in their able-bodied subject group. We hypothesize that the PEAs 25 observed in our study are less than those reported in [20] due to several differences in the experimental design and sub- ject population. First, our experimental design required the subject to stand within an instrumented walker, which lim- ited how far CoP could be adjusted in each direction, while subjects in the Lemay experiments stood with their hands at their sides without the constraint of an enclosure. Second, NO FO FR FL our experimental design also required that our subject Leaning posture direction adjusts the CoP by tracking a moving circle at a fixed velocity. Lemay’s subjects had 15 seconds to complete the leaning Figure 6: Mean maximum resultant UE force during leaning movements in the forward (FO), forward-right (FR), and forward- movement at a self-selected speed. Third, our subject was left (FL) directions. As a reference, the maximum resultant UE nonambulatory, had no control of his ankle plantar/dorsi- force exerted while the subject stood in the nominal (NO) starting flexors without stimulation, and required a support device position during a static trial is also displayed. Error bars are to stand. The subjects with SCI in the Lemay study were com- included to indicate ±1 standard deviation of measurements munity ambulators who could stand for 5 minutes without a across twelve repetitions. support device (AIS D), and many had near normal walking ability (1.02 m/s) [20]. Thus, they may not be representative changing posture, suggesting the impedance model may be a of individuals with incomplete SCI. Lemay et al. [20] further highly simplified representation of UE contribution to lean- state this as a possible reason that there was no statistical dif- ing movements. Thus, future work should explore more ference found in CoP between the SCI and able-bodied accurate representations of the interaction forces between area the UEs and the support device during leaning movements. groups. Future work will repeat these experiments with addi- Another potentially confounding assumption in the model tional subjects to determine PEAs that are more representa- was that the feet were fixed to the ground and not allowed tive of nonambulatory individuals with incomplete SCI. to rotate. Thus, the simulation outcomes could freely apply The CoP position feedback, as measured with force high-activation levels to the ankle plantar flexors without plates, was a practical signal for laboratory-based exploratory causing the model to fall over. These might have been the experiments with the PFC. However, the long-term goal is to reasons for a high reduction in the UE forces in simulation, deploy the controller for home use. Force plates limit con- which was not practicable in real life because of the troller deployment to the laboratory setting, but advances heel-lifting effect observed for the current study participant. in sensor technology enable the accurate capture of body PEA and ellipse width were computed to determine motion outside of a controlled laboratory environment. CoP-tracking deviations. Across all leaning directions, the Insole-pressure measurement devices are an appealing PEA increased as deviations in CoP tracking occurred option for the measurement of CoP, given that the position (Figure 7). The ellipse width provided an additional measure of the feet on the floor relative to each other are specified. of CoP tracking, as it described how far the overall CoP devi- Each time the user stands, it is likely that the location of the ated from the prescribed straight-line path (Figure 8). The feet will differ slightly. This is not a major issue in the labora- mean peak resultant UE effort, PEA, and ellipse width were tory, where the feet can be moved to fixed targets before each all greatest during leaning movements in the FL direction, experiment. However, for implementation in the uncon- but were relatively similar for postural changes in the FO trolled environments of the home and community, addi- and FR directions. These findings may suggest that as the tional sensors would need to be added to determine the leaning posture deviated from the prescribed path, more distances between the feet and their orientation before com- UE effort may have been required to readjust posture towards puting the CoP position. The CoM position is a global vari- the path. The PFC continually updated stimulation to the LEs able that can be implemented to detect the position of the as the changes in posture were elicited. The differences in the body each time the user stands as well as to track the dynamic findings for the FR and FL directions may be attributed to changes in posture as the user prepares for a functional task. differences in UE strength or the individual differences in Furthermore, the CoM position more accurately reflects the the stimulated responses of each muscle, among other issues system dynamics and can change without commensurate dis- particular to this subject. Future work will repeat similar placements of the CoP. The CoM position is therefore an experiments with additional subjects. ideal parameter for controlling the entire system, particularly Lemay et al. [20] conducted the comfortable multidirec- for faster movements or to recover from perturbations. tional limits of stability test with visual feedback to investi- Methods to estimate the CoM position from a network of gate dynamic postural stability in ambulatory individuals body-mounted inertial measurement units are underdevel- with SCI. They reported CoP ,defined by an ellipse fitting oped, and future work will verify such techniques and incor- area the linear distance between the initial and maximal positions porate them into home-going systems employing the of the CoP in each of the eight tested directions (FO, FR, FL, whole-body CoM position as the feedback signal. UE force (% BW) 10 Applied Bionics and Biomechanics 95% Prediction ellipse area for tracking tasks Posturogram of tracking tasks -10 −10 -20 −20 −50 −40 −30 −20 −10 0 10 20 30 40 50 -40 -20 0 20 40 60 CoP in ML direction (mm) CoP ML (mm) (a) (b) Figure 7: (a) Typical posturogram and (b) 95% prediction ellipses for CoP-tracking tasks in the forward and diagonal directions. The 95% 2 2 2 PEA for the leaning movements in the forward direction was 1276.5 mm , 1141.8 mm in the forward-right direction, and 1645.2 mm in the forward-left direction. Based on the orientation of the laboratory coordinate system, postural adjustments in the forward direction are indicated as CoP increasing from the nominal starting position. Postural shifts towards the left are indicated as CoP increasing from AP ML the nominal. Mean width of 95% prediction ellipses or coronal planes. Future work should explore and exploit the coupling between muscle actions and include cross terms 25 to represent the effects of the GS and TA on ML movement and PA and GMED on AP movement. This involves extend- ing the PFC to act in the generalized coronal plane and mod- ulating all muscles simultaneously irrespective of assumed movement direction (including the postural muscles for hip extension/flexion or trunk extension/lateral bending not adjusted in the current study) to generate the globally opti- mal patterns of stimulation to realize a movement. This study sought to determine the experimental feasibil- FO FR FL ity of the PFC, a muscle activation controller that modulated Leaning posture direction LE activation according to changes in the CoP position, in a recipient of an implanted standing NP. The PFC enabled Figure 8: Mean widths of 95% prediction ellipses for CoP-tracking the subject to assume leaning postures in the FO, FR, and tasks in the forward (FO), forward-right (FR), and forward-left (FL) FL directions, by modulating LE muscle activation according directions. The error bars are included to indicate ±1 standard to changes in the overall CoP position. More than twice the deviation of measurements across twelve repetitions. UE effort as a percentage of quiet standing were required to effect changes in CoP experimentally in this study as pre- dicted from the simulations presented in [10]. CoP-tracking A limitation of this study is the length of time the subject could stand during the experiments. Although the subject results indicate that all paths presented were successfully could stand quietly for 25 minutes at the time of testing, these tracked, suggesting that the PFC provided the subject with experiments were more demanding because they entailed more access to the workspace while standing. multiple repetitions of standing and adjusting posture in the different directions. To minimize fatigue induced by con- 5. Conclusions tinuous activation of the muscles, the number of repetitions collected was limited, so that the subject’s total standing time We have explored the experimental feasibility of the PFC, a did not exceed 10 minutes. This is consistent with elapsed CoP-position tracking muscle activation controller with a standing times with conventional FNS systems [2]. recipient of an implanted standing NP. This is the first study Another limitation to this study is the availability of mus- to our knowledge that investigates feedback control of stand- cles for control as well as the directions in which the recruited ing posture to enable user-selected leaning movements away muscles acted. The PFC, as implemented, assumed that the from erect stance in an individual with SCI. As the CoP posi- muscles acted independently and exclusively in the sagittal tion was adjusted to track the moving circle along the various Ellipse Width (mm) CoP in AP direction (mm) CoP AP (mm) Applied Bionics and Biomechanics 11 [8] H. Rouhani, M. Same, K. Masani, Y. Q. Li, and M. R. Popovic, paths, the PFC continually updated activation to the user’s “PID controller design for FES applied to ankle muscles in paralyzed LE musculature. Ellipse areas of the CoP traces neuroprosthesis for standing balance,” Frontiers in Neurosci- indicate that the PFC provided the user with greater access ence, vol. 11, p. 347, 2017. to the standing workspace. Future work will evaluate the con- [9] R. Nataraj, M. L. Audu, and R. J. Triolo, “Center of mass accel- troller with the whole-body CoM position as the feedback eration feedback control of standing balance by functional signal and account for cross-coupling resulting from the ana- neuromuscular stimulation against external postural perturba- tomical actions of the contracting muscles. This will require tions,” IEEE Transactions on Biomedical Engineering, vol. 60, the development and evaluation of a model that outputs no. 1, pp. 10–19, 2013. CoM from data captured from body-mounted sensors and [10] M. L. Audu, S. J. Gartman, R. Nataraj, and R. J. Triolo, “Postur- more advanced multidimensional control algorithms. e-dependent control of stimulation in standing neuroprosth- esis: simulation feasibility study,” Journal of Rehabilitation Data Availability Research and Development, vol. 51, no. 3, pp. 481–496, 2014. [11] S. J. Gartman, M. L. Audu, R. F. Kirsch, and R. J. Triolo, “Selec- The data used to support the findings of this study are tion of optimal muscle set for 16-channel standing neuro- available from the corresponding author upon request. prosthesis,” The Journal of Rehabilitation Research and Development, vol. 45, no. 7, pp. 1007–1018, 2008. Conflicts of Interest [12] J. J. Abbas and J. C. Gillette, “Using electrical stimulation to control standing posture,” IEEE Control Systems, vol. 21, The authors declare that there is no conflict of interest no. 4, pp. 80–90, 2001. regarding the publication of this paper. [13] N. Bhadra, K. L. Kilgore, and P. H. Peckham, “Implanted stim- ulators for restoration of function in spinal cord injury,” Med- ical Engineering & Physics, vol. 23, no. 1, pp. 19–28, 2001. Acknowledgments [14] M. A. Schiefer, K. H. Polasek, R. J. Triolo, G. C. J. Pinault, and This material is also the result of work supported with D. J. Tyler, “Selective stimulation of the human femoral nerve resources and the use of facilities at the Louis Stokes Cleveland with a flat interface nerve electrode,” Journal of Neural Engi- Veterans Affairs Medical Center in Cleveland, OH. Funding neering, vol. 7, no. 2, article 26006, p. 026006, 2010. for this work was provided by the National Institute of Neuro- [15] R. J. Triolo, C. Bieri, J. Uhlir, R. Kobetic, A. Scheiner, and E. B. logical Disorders and Stroke (Grant No: R01NS040547) and Marsolais, “Implanted functional neuromuscular stimulation the Craig H. Neilsen Foundation (Grant No: 459308). systems for individuals with cervical spinal cord injuries: clin- ical case reports,” Archives of Physical Medicine and Rehabili- tation, vol. 77, no. 11, pp. 1119–1128, 1996. References [16] B. Smith, Zhengnian Tang, M. W. Johnson et al., “An externally powered, multichannel, implantable stimulator- [1] G. P. Forrest, T. C. Smith, R. J. Triolo et al., “Use of the case telemeter for control of paralyzed muscle,” IEEE Transactions Western reserve/veterans administration neuroprosthesis for on Biomedical Engineering, vol. 45, no. 4, pp. 463–475, 1998. exercise, standing and transfers by a paraplegic subject,” Dis- ability and Rehabilitation: Assistive Technology, vol. 7, no. 4, [17] D. A. Winter, F. Prince, J. S. Frank, C. Powell, and K. F. Zabjek, pp. 340–344, 2012. “Unified theory regarding A/P and M/L balance in quiet stance,” Journal of Neurophysiology, vol. 75, no. 6, pp. 2334– [2] R. J. Triolo, S. N. Bailey, M. E. Miller et al., “Longitudinal per- 2343, 1996. formance of a surgically implanted neuroprosthesis for lower-extremity exercise, standing, and transfers after spinal [18] S. J. Gartman, M. L. Audu, R. F. Kirsch, and R. J. Triolo, “Selec- cord injury,” Archives of Physical Medicine and Rehabilitation, tion of optimal muscle set for 16-channel standing functional vol. 93, no. 5, pp. 896–904, 2012. electrical stimulation system,” Presented at the 12th Ann. [3] J. J. Abbas and H. J. Chizeck, “Feedback control of coronal Conf. Int. FES Soc, Philadelphia, PA, USA, 2007. plane hip angle in paraplegic subjects using functional neuro- [19] P. Schubert and M. Kirchner, “Ellipse area calculations and muscular stimulation,” IEEE Transactions on Biomedical Engi- their applicability in posturography,” Gait and Posture, neering, vol. 38, no. 7, pp. 687–698, 1991. vol. 39, no. 1, pp. 518–522, 2014. [4] M. Moynahan and H. J. Chizeck, “Characterization of paraple- [20] J.-F. Lemay, D. H. Gagnon, S. Nadeau, M. Grangeon, gic disturbance response during FNS standing,” IEEE Transac- C. Gauthier, and C. Duclos, “Center-of-pressure total trajec- tions on Rehabilitation Engineering, vol. 1, no. 1, pp. 43–48, tory length is a complementary measure to maximum excur- sion to better differentiate multidirectional standing limits of [5] R. J. Jaeger, “Design and simulation of closed-loop electrical stability between individuals with incomplete spinal cord stimulation orthoses for restoration of quiet standing in para- injury and able-bodied individuals,” Journal of NeuroEngineer- plegia,” Journal of Biomechanics, vol. 19, no. 10, pp. 825–835, ing and Rehabilitation, vol. 11, no. 1, p. 8, 2014. [6] K. J. Hunt, H. Gollee, and R. P. Jaime, “Control of paraplegic ankle joint stiffness using FES while standing,” Medical Engi- neering & Physics, vol. 23, no. 8, pp. 541–555, 2001. [7] Z. Matjacic and T. Bajd, “Arm-free paraplegic standing- part II: experimental results,” IEEE Transactions on Rehabilitation Engineering, vol. 6, no. 2, pp. 139–150, 1998. International Journal of Advances in Rotating Machinery Multimedia Journal of The Scientific Journal of Engineering World Journal Sensors Hindawi Hindawi Publishing Corporation Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 http://www www.hindawi.com .hindawi.com V Volume 2018 olume 2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Journal of Control Science and Engineering Advances in Civil Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 Submit your manuscripts at www.hindawi.com Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 VLSI Design Advances in OptoElectronics International Journal of Modelling & Aerospace International Journal of Simulation Navigation and in Engineering Engineering Observation Hindawi Hindawi Hindawi Hindawi Volume 2018 Volume 2018 Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com www.hindawi.com www.hindawi.com Volume 2018 International Journal of Active and Passive International Journal of Antennas and Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration Hindawi Hindawi Hindawi Hindawi Hindawi www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal

Applied Bionics and BiomechanicsHindawi Publishing Corporation

Published: Mar 14, 2019

References