Forelimb EMG-based trigger to control an electronic spinal bridge to enable hindlimb stepping after a complete spinal cord lesion in rats
- Parag Gad†1,
- Jonathan Woodbridge†2,
- Igor Lavrov†3,
- Hui Zhong3,
- Roland R Roy3, 5,
- Majid Sarrafzadeh2 and
- V Reggie Edgerton3, 4, 5Email author
© Gad et al.; licensee BioMed Central Ltd. 2012
Received: 20 July 2011
Accepted: 20 April 2012
Published: 12 June 2012
A complete spinal cord transection results in loss of all supraspinal motor control below the level of the injury. The neural circuitry in the lumbosacral spinal cord, however, can generate locomotor patterns in the hindlimbs of rats and cats with the aid of motor training, epidural stimulation and/or administration of monoaminergic agonists. We hypothesized that there are patterns of EMG signals from the forelimbs during quadrupedal locomotion that uniquely represent a signal for the “intent” to step with the hindlimbs. These observations led us to determine whether this type of “indirect” volitional control of stepping can be achieved after a complete spinal cord injury. The objective of this study was to develop an electronic bridge across the lesion of the spinal cord to facilitate hindlimb stepping after a complete mid-thoracic spinal cord injury in adult rats.
We developed an electronic spinal bridge that can detect specific patterns of EMG activity from the forelimb muscles to initiate electrical-enabling motor control ( eEmc) of the lumbosacral spinal cord to enable quadrupedal stepping after a complete spinal cord transection in rats. A moving window detection algorithm was implemented in a small microprocessor to detect biceps brachii EMG activity bilaterally that then was used to initiate and terminate epidural stimulation in the lumbosacral spinal cord. We found dominant frequencies of 180–220 Hz in the EMG of the forelimb muscles during active periods, whereas these frequencies were between 0–10 Hz when the muscles were inactive.
Results and conclusions
Once the algorithm was validated to represent kinematically appropriate quadrupedal stepping, we observed that the algorithm could reliably detect, initiate, and facilitate stepping under different pharmacological conditions and at various treadmill speeds.
KeywordsSpinal cord injury Spinal bridge-assisted stepping EMG detection Fast Fourier transform
Functionally complete spinal cord injury is a severe debilitating condition and leads to paralysis. Numerous approaches have been attempted to recover function after paralysis, e.g., facilitation of axon regeneration including methods to suppress growth inhibitory molecules, modulation of the levels of neurotrophic factors, cell transplantation, and the use of activity-dependent mechanisms[1–4]. These techniques, however, have not resulted in dramatic improvements of motor function after motor complete paralysis. A technique that has shown promise is Brain-Computer Interface. This approach has been successfully developed in integrating activity from the functionally unaffected sites, such as the motor cortex, to control robotic devices or muscle stimulating devices to generate the desired movement in paralyzed muscle groups.
Recent in vivo studies in rats and cats show that networks of neurons in the lumbosacral region of the spinal cord have an intrinsic capability to generate coordinated rhythmic motor outputs in the hindlimbs[2, 6, 7]. Several strategies have been tested to tap into these neural circuits and activate them to induce oscillatory motions in the hindlimbs. For example, pharmacologically enabling motor control strategies (fEmc) using serotonergic agonists of 5-HT1A,2A and 5-HT7 receptors in combination with epidural stimulation, i.e., electrical-enabling motor control (eEmc)[8–10], have been used to recover considerable function after paralysis. These two interventions combined with the availability of sensory information in real time have been used to induce full weight-bearing stepping in complete spinal rats[10, 11]. Gerasimenko et al. have shown that eEmc (at 40 Hz with monopolar stimulation) between the L2 and S1 spinal cord levels facilitates bilateral stepping of spinal rats on a moving treadmill belt. In contrast, the spinal rats did not step when the treadmill was turned on but no eEmc was provided, indicating that the stimulation was necessary for the spinal rats to step.
These observations led us to ask whether ‘indirect’ volitional control of eEmc (by forelimb EMG activity) could be used to facilitate stepping in the hindlimbs and provide a new and stable level of control of motor function in spinal rats. Therefore, the purpose of this study was to determine whether ‘indirect’ volitional control via an electronic spinal bridge could be accomplished. In human subjects, this volitional control could avoid the use of an external switch to activate an electrode array by using EMG signals as occurs during normal locomotion. As importantly the present experiments provide a testbed for development of a Brain-Machine-Spinal Cord Interface (BMSCI) that allows for motor control with minimal conscious attention. In addition, it provides a potential mechanism for exerting finer motor control than could be accomplished using a simple on/off system. To design and test such a ‘Brain-Machine-Spinal Cord Interface’, we used EMG signals from the forelimbs as a trigger to initiate spinal cord stimulation to facilitate movement of the hindlimbs. We used the forelimb EMG because these signals are part of the natural gait cycle in quadrupedal stepping.
The objectives of this study were to develop an effective pattern of step detection from the uninjured forelimb muscles and to use this pattern to control the on/off state of the eEmc of the spinal cord to facilitate quadrupedal stepping under different experimental conditions in spinal rats. We observed that voluntary signals from the forelimbs during stepping can be used as a control mechanism for generating signals to the spinal cord to facilitate hindlimb stepping. The underlying assumption is that when the rat “intends” to step the forelimbs will be activated in a pattern that reflects this “voluntary” intent and that intent will initiate eEmc to facilitate stepping of the hindlimbs.
Adult female Sprague–Dawley rats (n=5, ~300 g body weight) were used. Pre- and post-surgical animal care has been described previously. The rats were housed individually with food and water provided ad libitum. All survival surgical procedures were conducted under aseptic conditions and with the rats deeply anesthetized (isoflurane gas administered via facemask as needed). All procedures described below are in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals and were approved by the Animal Research Committee at UCLA.
The rats underwent two separate surgeries. The first surgery was to implant the EMG electrodes. The rats were allowed to recover from this implant surgery for one week and then recordings (pre-transection) were made while the rats stepped on the treadmill. After completing these recordings, the rats underwent a second surgery during which the spinal cord was completely transected at a mid-thoracic level and epidural electrodes were implanted at spinal levels L2 and S1. The rats were allowed to recover for one week and then the training sessions were initiated. Recordings were performed to test the electronic bridge at 5 weeks post-transection. The pre-transection EMG recordings provided a baseline for comparison with the recordings post-transection. All of these procedures are performed routinely in our laboratory[9, 13]. Details of each step are given below.
Head connector implantation
A small incision was made at the midline of the skull. The muscles and fascia were retracted laterally, small grooves were made in the skull with a scalpel, and the skull was dried thoroughly. Two amphenol head connectors with Teflon-coated stainless steel wires (AS632, Cooner Wire, Chatsworth CA) were securely attached to the skull with screws and dental cement as described previously[14, 15].
Intramuscular EMG electrode implantation
Selected hindlimb (tibialis anterior, TA; and soleus, Sol) and forelimb (biceps brachii, BB; and triceps brachii, TB) muscles were implanted bilaterally with EMG recording electrodes as described by Roy et al.. Skin and fascial incisions were made to expose the belly of each muscle. Two wires extending from the skull-mounted connector were routed subcutaneously to each muscle. The wires were inserted into the muscle belly using a 23-gauge needle and a small notch (~0.5-1.0 mm) was removed from the insulation of each wire to expose the conductor and form the electrodes. The wires were secured in the belly of the muscle via a suture on the wire at its entrance into and exit from the muscle belly. The wires were looped at the entrance site to provide stress relief. The proper placement of the electrodes was verified during the surgery by stimulating through the head connector and post-mortem via dissection.
Spinal cord transection
A partial laminectomy was performed at the T8-T9 vertebral level and a longitudinal cut was made in the dura to expose the spinal cord. A complete spinal cord transection to include the dura was performed at approximately the T8 spinal level using microscissors. Two surgeons verified the completeness of the transection by lifting the cut ends of the spinal cord and passing a glass probe through the lesion site. Gel foam was inserted into the gap created by the transection as a coagulant and to separate the cut ends of the spinal cord.
Epidural electrode implantation
Epidural electrodes were coiled and left in the back region of the animal after the EMG surgery. The epidural electrodes were implanted during the second surgery. Partial laminectomies were performed to expose the spinal cord at spinal levels L2 and S1. Two Teflon-coated stainless steel wires from the head connector were passed under the spinous processes and above the dura mater of the remaining vertebrae between the partial laminectomy sites. After removing a small portion (~1 mm notch) of the Teflon coating and exposing the wire on the surface facing the spinal cord, the electrodes were sutured to the dura mater at the midline of the spinal cord above and below the electrode sites using 8.0 Ethilon suture ( Ethicon, New Brunswick, NJ). A common ground (indifferent) wire (~1 cm of the Teflon coating removed distally) was inserted subcutaneously in the mid-back region. All wires were coiled in the back region to provide stress relief.
All incision areas were irrigated liberally with warm, sterile saline. All surgical sites were closed in layers, i.e., muscle and connective tissue layers with Vicryl (Ethicon, New Brunswick, NJ) and the skin incision on the back with Ethilon and in the limbs with Vicryl. All closed incision sites were cleansed thoroughly with saline solution. Analgesia was provided by buprenex (0.5–1.0 mg/kg, s.c., 3 times/day). The analgesics were initiated before completion of the surgery and continued for a minimum of 2 days. The rats were allowed to fully recover from anesthesia in an incubator. The rats were housed individually, and the bladders of the spinal rats were expressed manually 3 times/day for the first 2 weeks after surgery and 2 times per day thereafter. The hindlimbs of the spinal rats were moved passively through a full range of motion once per day to maintain joint mobility. All of these animal care procedures have been described in detail previously.
Stimulation and training procedures
All rats were trained to step quadrupedally using a body weight support system under the influence of quipazine administration (0.3 mg/kg, i.p.) and eEmc (40 Hz, between L2 and S1 with the current flowing from L2 to S1)[9, 13, 16]. The maximum stimulation voltage used was 3 V and the stimulation intensity was modulated to produce maximum stepping performance. The rats stepped on a specially designed motor-driven rodent treadmill. The treadmill belt had an anti-slip material that minimized slipping while stepping. The rats were trained using a body weight support system: the rats were suspended in a jacket such that all four limbs were in contact with the treadmill and that there was enough room for all 4 limbs to carry out the swing and stance phases of the step cycle. This was a critical component of the design as it was important to engage the forelimbs in stepping to produce robust, high quality EMG signals from the forelimb muscles.
Pre-transection the rats were stepped quadrupedally on the treadmill at varying speeds (13.5 to 21 cm/s) without the use of the body weight support system. These baseline recordings were compared to post-transection recordings. Five days post-transection, the rats were fitted with a jacket and secured to the body weight system for a period of 2–3 min initially and then the time was progressively increased to about 10 min by day 7. This was an acclimation period and the rats were not stepped during this period. Training began one-week post-transection. Stepping ability was tested once a week pre-quipazine and 15 minutes post-quipazine administration. Quipazine (a serotoninergic agonist) administered intraperitoneally (0.3 mg/kg) has been shown to improve stepping performance of spinal animals when receiving eEmc[8, 16]. We used quipazine administration in the present study to produce robust stepping in the spinal rats. Kinematics and EMG data were collected on a weekly basis from all rats. The algorithm for detection of forelimb stepping was based on these data.
Data acquisition and post-processing
Kinematics recording parameters
Addtitional file 1:The video file demonstrates a spinal rat stepping quadrupedally on a treadmill at 13.5 cm/s under the influence of eEmc (40 Hz between L2 and S1 spinal levels) and quipazine (0.3 mg/kg, i.p.). The file can be viewed using any media player such as vlc or windows media player. (MPEG 3 MB)
Electronic bridge schematic
Figure1A shows the schematic of the electronic bridge and the experimental setup. The EMG signals from the forelimb muscles were fed to the electronic bridge. The output of the electronic bridge was connected to wire electrodes implanted at specific levels on the spinal cord. Figure1B shows an expanded view of the electronic bridge. We used an 8:1 MUX (MAX14752, Maxim) that is controlled by a microcontroller (MSP EZ430, Texas Instruments). The electronic bridge has 2 input channels (RBB and LBB) and one output channel (Stim 1) while 5 other channels are reserved for future use. The EZ430 contains 10 I/O channels and 5 of these channels are used for this system (3 control lines, 1 input and 1 output).
The MSP EZ430 has an inbuilt 10-bit ADC with a maximum sampling rate of 200 kHz. It consists of 10 general-purpose analog I/O lines that allow the design to be flexible along with potential additions to the design in the future. The MSP EZ430 is powered by a DC source that provides a maximum output of 3.3 V. The minimum voltage required to trigger the Grass stimulator is 5 V. The 3.3 V output is converted to a 5 V output using a simple NE555 timer that is synchronized with the output of the MSP EZ430 to provide 40 Hz pulses.
EMG detection techniques
Figure4 represents the result of the frequency analysis technique. We observed that the RBB has higher baseline noise as compared to the LBB. Using the frequency response we were able to successfully define the bursts even in the case of large baseline noise. The use of the second technique allowed us to detect stepping based on the variation in the EMG frequency spectrum between the active phase and the inactive phase independent of the amplitude of the EMG signals.
We tested the stability of the detection algorithm of the bridge 5 weeks post-transection at treadmill speeds of 13.5 and 21 cm/s and with and without quipazine administration. This enabled us to test the stability of the detection algorithm under different conditions (better hindlimb stepping with than without quipazine administration at both speeds). All rats were tested for a total of 5 trials under each condition and the 3 best trials were chosen for analysis. The efficiency of the electronic bridge was tested by measuring ton and toff for the various test conditions. ton was defined as the time between the treadmill turning on and the stimulation turning on (time needed for the bridge to detect stepping). toff was defined as the time between the treadmill turning off and the stimulation turning off (time needed for the bridge to detect the end of stepping).
We also tested the stepping ability of the rats at two treadmill speeds with and without quipazine with direct stimulation (without the electronic bridge) where the experimenter turned on the stimulation manually. This enabled us to compare the stepping ability with and without the electronic bridge.
Having identified three criteria based on FFT analysis of EMG as described above it was also necessary to validate whether these criteria resulted in kinematics characteristics associated with and without forelimb weight bearing. This validation was performed with respect to the speed of locomotion and with different pharmacological interventions. The single kinematics criteria reflecting successful stepping was a change in elbow angle bilaterally of at least 50° for five consecutive steps.
All data are reported as mean±SEM. Statistically significant differences were determined using a two-way (ton and toff under 4 experimental conditions) analysis of variance (ANOVA). The criterion level for determination of statistical significance was set at P < 0.05 for all computations.
Voluntarily induced stepping
Does the electronic bridge have a detrimental effect on stepping performance?
Algorithm validation: Determining the detection threshold for stepping
We have developed a BMSCI having a pattern recognition algorithm that can use EMG from the forelimb muscles to trigger the initiation and termination of the stimulation of the spinal cord below the level of a complete spinal cord injury. This algorithm (second generation) detects stepping with little or no calibration and thus provides an advantage over the system (first generation) we tested initially that needs constant monitoring. Our logic for using the EMG signals from the forelimb muscles as a trigger was that these signals reflect the "intent to step" quadrupedally. The idea was to use naturally generated EMG signals from the forelimbs to control an electronic bridge that would facilitate hindlimb stepping in spinal rats. Our results from multiple test conditions demonstrate that this system is capable of adapting to different pharmacological and stepping conditions. Results from the first generation showed us that it is possible to develop a real time EMG detection system on the MSP430, but this generation was cumbersome to run due to calibration of the multiple channels and the variability among animals. The technique used in the second generation allowed us to reduce the calibration and to accommodate the variability among animals. Given the variability in the conditions under which the step detection algorithm was tested on the MSP430, the time taken to detect the beginning of stimulation (ton) and the time taken to detect the stopping of stimulation (toff) had about the same consistency and demonstrated the algorithm's ability to detect EMG signals and to adapt to different conditions and different animals (Figure6).
To further enhance the utility of spinal cord stimulation, the control system must go beyond an on/off control as shown in the present experiments. In paraplegic humans any set of muscles that can be voluntarily controlled could be used as a trigger. Generating and controlling the EMG in muscles such as the deltoid or pectoralis to control prosthetic devices has been shown to be feasible in humans. Harkema et al. demonstrated in a single human subject the importance of varying stimulation parameters in obtaining the most effective standing and voluntary control of the legs in a subject with a motor complete injury. Similarly, Ichiyama et al. reported that there was variation in the required stimulation voltage at different time points after injury: the stimulation intensity needed to generate locomotor activity in spinal rats was 3.39 ± 0.2 V at 2 weeks and 7.55 ± 1.2 V at 4 weeks post-ST using 40 Hz stimulation monopolar stimulation at L2.
Dutta et al. reported a system to detect EMG during stepping to trigger muscles in the contralateral limb to initiate stepping in human patients with an incomplete spinal cord injury. A training data set was derived from patients with a switch-based functional electrical stimulation system to assist the patients in stepping and a feature extraction was performed on the EMG. A threshold-based binary classifier was trained to distinguish a set of feature templates in the EMG linear envelopes indicating the intention to trigger the next step. These authors report that during online testing the intention to trigger the next step was detected using analyses between features extracted from real time EMG linear envelopes and feature templates derived from the training data set.
The ultimate objective of the present study was to use EMG from multiple muscles in the forelimb to detect stepping and fine variations in forelimb and hindlimb stepping and to modulate the stimulation of the spinal cord to generate optimum stepping movements in the hindlimbs. We also have developed a simulated test bench in MATLAB (Mathworks) to test the EMG signals for offline step detection. This testbench is an exact replica of the algorithm run on the MSP430. Using this simulated test bench, we can validate the recordings seen in real time as well as test the detection algorithm at different time points during recovery after an injury. Further simulations could be used to develop a ‘self-learning’ or ‘threshold-adjusting’ algorithm that could be animal specific as well as being applicable for different stages of recovery after an injury.
We have developed a novel technique for detecting stepping of the forelimbs and neuromodulating the spinal circuitry in real time to control hindlimb movements in rats with complete paralysis. This detection algorithm can accommodate the variations in EMG amplitudes that normally occur during spontaneous functional recovery after a spinal cord injury. This neuromodulatory approach also is likely to have the potential to improve the control of movements in other neuromotor disorders, such as stroke and Parkinson Disease.
Brain-Machine-Spinal Cord Interface
Discrete Fourier Transform
Left Biceps Brachii
Fast Fourier Transform
Left Tibialis Anterior
Left Triceps Brachii
Right Biceps Brachii
Right Tibialis Anterior
Right Triceps Brachii.
We would like to thank Maynor Herrera for providing excellent animal care and Sharon Zdunowski for technical assistance. We also would like to thank the undergraduate students, i.e., Jacquin Galazara, Eric Kim, Andrew Chon and James Song who helped us in performing the experiments and analyzing the data.
This research was supported by the National Institute of Biomedical Imaging and Bioengineering R01EB007615, the Christopher and Dana Reeve Foundation, the Roman Reed Spinal Cord Injury Research Fund of California and the Russian Foundation for Basic Research 11-04-12074-OFi-M-2011.
- Thuret S, Moon L, Gage F: Therapeutic interventions after spinal cord injury. Nature 2006, 7: 628-643.Google Scholar
- Edgerton VR, Tillakaratne NJT, Bigbee AJ, de Leon RD, Roy RR: Plasticity of the spinal circuitry after injury. Ann Rev Neurosci 2004, 27: 145-167. 10.1146/annurev.neuro.27.070203.144308View ArticlePubMedGoogle Scholar
- Edgerton VR, Courtine G, Gerasimenko Y, Lavrov I, Ichiyama RM, Fong AJ, Cai LL, Otoshi CK, Tillakaratne N, Burdick JW, Roy RR: Training locomotor networks. Brain Res Rev 2008, 57: 241-254. 10.1016/j.brainresrev.2007.09.002PubMed CentralView ArticlePubMedGoogle Scholar
- Lee Y, Sindhu R, Lin C, Ehdaie A, Lin V, Vaziri N: Effects of nerve graft on nitric oxide synthase, NAD(P)H oxidase, and antioxidant enzymes in chronic spinal cord injury. Free Radical Biol 2004, 36: 330-339. 10.1016/j.freeradbiomed.2003.11.006View ArticleGoogle Scholar
- Muller G, Pfurtscheller J, Jurgen Gerner H, Rupp R: Thought - Control of functional electrical stimulation to restore hand grasp in a patient with tetraplegia. Neurosci Lett 2003, 351: 33-36. 10.1016/S0304-3940(03)00947-9View ArticlePubMedGoogle Scholar
- Ivashko DG, Prilutsky BI, Markin SN, Chapin JK, Rybak IA: Modeling the spinal cord neural circuitry controlling cat hindlimb movement during locomotion. Neurocomputing 2003, 52–54: 621-629.View ArticleGoogle Scholar
- Gerasimenko YP, Avelev VD, Nikitin OA, Lavrov IA: Initiation of locomotor activity in spinal cats by epidural stimulation of the spinal cord. Neurosci Behav Physiol 2003, 33: 247-254. 10.1023/A:1022199214515View ArticlePubMedGoogle Scholar
- Gerasimenko YP, Ichiyama RM, Lavrov IA, Courtine G, Cai L, Zhong H, Roy RR, Edgerton VR: Epidural spinal cord stimulation plus quipazine administration enable stepping in complete spinal adult rats. J Neurophysiol 2007, 98: 2525-2536. 10.1152/jn.00836.2007View ArticlePubMedGoogle Scholar
- Courtine G, Gerasimenko Y, van den Brand R, Yew A, Musienko P, Zhong H, Song B, Ao Y, Ichiyama RM, Lavrov I, Roy RR, Sofroniew MV, Edgerton VR: Transformation of nonfunctional spinal circuits into functional states after the loss of brain input. Nat Neurosci 2009, 12: 1333-1342. 10.1038/nn.2401PubMed CentralView ArticlePubMedGoogle Scholar
- Goldberger ME: Spared root deafferentation of a cat's hindlimb: hierachial regulations of pathways mediating recovery of motor behavior. Exp Brain Res 1988, 73: 329-342.PubMedGoogle Scholar
- Lavrov I, et al.: Epidural stimulation induced modulation of spinal locomotor networks in adult spinal rats. J Neurosci 2008, 28: 6022-6029. 10.1523/JNEUROSCI.0080-08.2008PubMed CentralView ArticlePubMedGoogle Scholar
- Roy RR, Hodgson JA, Lauretz SD, Pierotti DJ, Gayek RJ, Edgerton VR: Chronic spinal cord-injured cats: surgical procedures and management. Lab Anim Sci 1992, 42: 335-343.PubMedGoogle Scholar
- Gerasimenko YP, Ichiyama RM, Zhong H, Roy RR, Edgerton VR: Features of bipedal stepping induced by epidural spinal cord stimulation and quipazine administration in spinal rats. International Society for Posture and Gait Research 2005-XVII Conference, Marseille, France; 2005. May 29-June 2Google Scholar
- Roy RR, Hutchison DL, Pierotti DJ, Hodgson JA, Edgerton VR: EMG patterns of rat ankle extensors and flexors during treadmill locomotion and swimming. J Appl Physiol 1991, 70: 2522-2529. 11PubMedGoogle Scholar
- Ichiyama RM, Gerasimenko YP, Zhong H, Roy RR, Edgerton VR: Hindlimb stepping movements in complete spinal rats induced by epidural spinal cord stimulation. Neurosci Lett 2005, 383: 339-344. 12 10.1016/j.neulet.2005.04.049View ArticlePubMedGoogle Scholar
- Ichiyama RM, Gerasimenko Y, Jindrich DL, Zhong H, Roy RR, Edgerton VR: Dose dependence of the 5-HT agonist quipazine in facilitating spinal stepping in the rat with epidural stimulation. Neurosci Lett 2008, 438: 281-285. 10.1016/j.neulet.2008.04.080PubMed CentralView ArticlePubMedGoogle Scholar
- Jacobsen E, Lyons R: The sliding DFT. IEEE Signal Process Mag Mar. 2003, 20: 74-80.Google Scholar
- Cooley JW, Lewis P, Welch P: The fast fourier transform and its applications. IEEE Trans on Education 1969,12(1):28-34.View ArticleGoogle Scholar
- Zecca M, Micera S, Carrozza M, Dario P: Control of multifunctional prosthetic hands by processing the electromyographic signal. Crit Rev Biomed Eng 2002, 30: 459-485. 10.1615/CritRevBiomedEng.v30.i456.80View ArticlePubMedGoogle Scholar
- Harkema S, Gerasimenko YP, Hodes J, Burdick J, Angeli C, Chen Y, Ferreira C, Willhite A, Rejc E, Grossman R, Edgerton VR: Effect of epidural stimulation of the lumbosacral spinal cord on voluntary movement, standing and assisted stepping after motor complete paraplegia: a case study. Lancet 2011, 377: 1938-1947. 10.1016/S0140-6736(11)60547-3PubMed CentralView ArticlePubMedGoogle Scholar
- Dutta A, Kobetic R, Triolo R: Gait Initiation with electromyographically triggered electrical stimulation in people with partial paralysis. J Biomech Eng 2009, 131: 081001-081009. 10.1115/1.3127248View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License(http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.