The Evoked Potential Operant Conditioning System aids scientific investigation of sensorimotor function and can administer targeted neurobehavioral training that can impact sensorimotor rehabilitation in neuromuscular disorders. This article describes its capabilities and illustrates its application in modifying a simple spinal reflex to achieve lasting improvement in motor function.
The Evoked Potential Operant Conditioning System (EPOCS) is a software tool that implements protocols for operantly conditioning stimulus-triggered muscle responses in people with neuromuscular disorders, which in turn can improve sensorimotor function when applied appropriately. EPOCS monitors the state of specific target muscles—e.g., from surface electromyography (EMG) while standing, or from gait cycle measurements while walking on a treadmill—and automatically triggers calibrated stimulation when pre-defined conditions are met. It provides two forms of feedback that enable a person to learn to modulate the targeted pathway's excitability. First, it continuously monitors ongoing EMG activity in the target muscle, guiding the person to produce a consistent level of activity suitable for conditioning. Second, it provides immediate feedback of the response size following each stimulation and indicates whether it has reached the target value.
To illustrate its use, this article describes a protocol through which a person can learn to decrease the size of the Hoffmann reflex—the electrically-elicited analog of the spinal stretch reflex—in the soleus muscle. Down-conditioning this pathway's excitability can improve walking in people with spastic gait due to incomplete spinal cord injury. The article demonstrates how to set up the equipment; how to place stimulating and recording electrodes; and how to use the free software to optimize electrode placement, measure the recruitment curve of direct motor and reflex responses, measure the response without operant conditioning, condition the reflex, and analyze the resulting data. It illustrates how the reflex changes over multiple sessions and how walking improves. It also discusses how the system can be applied to other kinds of evoked responses and to other kinds of stimulation, e.g., motor evoked potentials to transcranial magnetic stimulation; how it can address various clinical problems; and how it can support research studies of sensorimotor function in health and disease.
Over the past decade, targeted neuroplasticity strategies have emerged as a new approach for the rehabilitation of neurological impairments1,2. One such strategy is operant conditioning of an evoked potential. This entails repeatedly eliciting electrophysiological responses that can be measured non-invasively – for example, by electroencephalography (EEG) or surface electromyography (EMG) – and giving the person immediate feedback on the size of each response relative to a criterion level set by the therapist or investigator. Over time, this protocol trains the person to increase or decrease their response and can, consequently, target beneficial change to a central nervous system site that is important in a behavior such as locomotion or reach-and-grasp. The targeted change benefits performance and, in addition, enables better practice that leads to widespread beneficial change that improves the entire behavior. For example, in people with incomplete spinal cord injury (iSCI) in whom clonus impairs locomotion, operant conditioning that reduces the Hoffmann reflex in the soleus muscle of one leg improves locomotor muscle activity in both legs, thereby increasing walking speed and restoring right/left step symmetry1,3,4,5. Another example is that of paired-pulse stimulation, which can lastingly increase the size of the motor-evoked potential (MEP) to transcranial magnetic stimulation, thereby improving reach-and-grasp function in people with chronic hand and arm impairment following iSCI6.
Implementing such protocols demands special-purpose software that must perform multiple functions. Specifically, it must continuously acquire, process, and save electrophysiological signals; it must continuously monitor the state of the nervous system and trigger stimulation appropriately under tight real-time constraints; it must provide continuous moment-by-moment feedback, trial-by-trial feedback, and session-by-session feedback; it must provide a user interface to guide setup and tuning by the investigator or therapist; and, finally, it must store and organize signal data and meta-information in a standardized format.
The evoked potential operant conditioning system (EPOCS) is our answer to this outstanding need. Under the hood, the software is based on BCI2000, an open-source neurotechnology platform that is used in hundreds of laboratories around the world7,8. In EPOCS, BCI2000's usual user interface is hidden and replaced by a streamlined interface that is optimized for evoked potential operant conditioning protocols.
The current article and its accompanying video illustrate the use of EPOCS in one particular protocol: operant conditioning to reduce the size of the Hoffmann (H-) reflex in the soleus muscle. This response is the electrically elicited analog of the knee-jerk stretch reflex. H-reflex down-conditioning has been shown to reduce the impact of clonus on, and to thereby improve, locomotion in animals with iSCI9,10,11,12,13 and in humans with iSCI, multiple sclerosis, or stroke5,14,15. It can be applied without adverse side effects in animals and people with or without neurological injury16,17.
The operant conditioning protocol functions by performing multiple trials, each lasting several seconds. The sequence of events of one trial is shown schematically in Figure 1, with numbers denoting the following functions:
1. Continuous background EMG is recorded from bipolar surface electrodes over the target muscle (soleus) and its antagonist (tibialis anterior). The background level is evaluated as the mean rectified value of the high-pass-filtered signal in a sliding window.
2. Background EMG level in the target muscle is shown as the height of a bar, continuously updated on the participant's screen. This helps the participant to keep the activity within a specified range (hatched region).
3. The software judges the appropriate moment for electrical stimulation and triggers the stimulator accordingly. The principal criteria are that at least 5 s must have elapsed since the previous stimulation and that the background EMG level must have remained in the specified range continuously for 2 s.
4. A constant-current stimulator delivers an electrical pulse transcutaneously to the tibial nerve (typically monophasic, with 1 ms duration).
5. The resulting stimulus-locked response is recorded. The software computes the sizes of two components of particular interest: the earlier M-wave, which reflects muscle activation resulting from direct stimulation of the motor axon; and the later H-reflex, which reflects the signal relayed through a reflex arc in the spinal cord18,19,20,21,22. EPOCS refers to these as the reference response and target response, respectively.
6. H-reflex size for the current trial is displayed as the height of a second bar, relative to a desired criterion level that defines a successful or unsuccessful trial. For down-conditioning, the bar is dark green if the H-reflex size fell below the criterion, or bright red if it did not (vice versa for up-conditioning). Simultaneously, the numeric display of the cumulative success rate is updated accordingly. Together, these graphical display elements provide the immediate positive or negative reinforcement on which operant conditioning relies23.
Figure 1: Schematic illustration of EPOCS' core functionality during down-conditioning of the soleus H-reflex. The participant views a large monitor screen that shows the background EMG level, the most recent H-reflex size, the number of trials completed so far in the current run of 75, and the running proportion of successful trials for the run. The sequence of events in one trial is denoted by the numbers 1-6, as described in the Introduction. Please click here to view a larger version of this figure.
A human H-reflex conditioning protocol typically consists of 6 baseline sessions, followed by 24-30 conditioning sessions spread over 10 weeks at a rate of 3 sessions/week, and several follow-up sessions over the subsequent 3-6 months14,16. Each session lasts 60-90 min.
To support this protocol as well as other related protocols, EPOCS has five distinct modes of operation, each served by one of the tabs of its main window, entitled Stimulus Test, Voluntary Contraction, Recruitment Curve, Control Trials, and Training Trials.
In Stimulus Test mode, the software triggers a stimulus every few seconds, not necessarily contingent on the state of the target muscle. The response signals are shown on the screen after each stimulus. This allows the operator to verify the quality of the electrode connections and the EMG signal; to optimize the position of the stimulating and recording electrodes; and to establish the individual's response morphology.
In Voluntary Contraction mode, the software measures and shows the background EMG level while the participant is encouraged to contract the muscle as much as possible, in the absence of electrical stimulation. In some protocols, the EMG level at maximum voluntary contraction (MVC) is a useful reference for setting the background EMG criteria. In the protocol demonstrated here, this is not necessary, as a stable standing posture standardizes the activity of the soleus muscle sufficiently.
In Recruitment Curve mode, stimulation is contingent on the background EMG level (shown continuously on the screen) remaining in the correct range; response signals are shown on the screen after each stimulus; and the sequence of responses may be analyzed at the end of a run. This allows the operator to determine the start and end of the time intervals in which the responses of interest appear; to determine the relationship between stimulation intensity and response size, both before and after the conditioning runs; and to determine the stimulation intensity to be used for conditioning.
In Control Trials mode, stimulation is contingent on the background EMG level (shown continuously on screen), but no feedback is given about the target response size. The sequence and distribution of response sizes may be analyzed. This mode may be used to gather baseline measurements of response size, or as a control condition for comparison against operant conditioning in a crossover or between-subjects experimental design. It can serve as a basis for setting the performance criterion for operant conditioning at the beginning of each session.
Finally, in Training Trials mode, stimulation is contingent on the background EMG level (shown continuously on screen), and trial-by-trial reinforcement is also provided by showing the target response size, as described above and shown in Figure 1. This is the mode in which operant conditioning is performed.
The next section will guide the reader through the five modes by demonstrating the protocol for down-conditioning the soleus H-reflex in an adult participant without neurological injury.
All procedures described here were approved by the institutional review board of the Stratton VA Medical Center (approval number 1584762-9). The participant in the video gave informed consent for the use of their image and EMG signals in this publication.
NOTE: The terms in bold indicate labels that should be visible on the hardware and/or in the software graphical user interface.
1. Software setup
2. Hardware setup
3. Preparing stimulation and recording electrodes
4. Using the EPOCS software
5. Performing multiple repeat sessions
Figure 2 shows the effectiveness of the above protocol in measuring M-wave and H-reflex recruitment curves and in measuring the distribution of H-reflex sizes at constant stimulation intensity. It also illustrates the overall effectiveness of the multi-session protocol in changing the H-reflex size in neurologically unimpaired participants and in improving locomotor function in participants with incomplete spinal cord injury.
Figure 2A shows a screenshot of the analysis window following a run performed in Recruitment Curve mode during H-reflex operant conditioning (see protocol step 4.5.). In the lower half of the window (the Sequence pane), the horizontal axis shows trial number—hence, stimulus intensity increases from left to right. H-reflex size (green circles) rises then falls as a function of stimulus intensity, whereas M-wave size (brown triangles) rises then saturates. Figure 2B shows a screenshot of the analysis window following a run performed in Control Trials or Training Trials mode during H-reflex operant conditioning (see protocol step 4.6. and step 4.7.). In the lower panel (the "Distribution" pane), the histogram of H-reflex sizes facilitates the selection of an appropriate criterion level for subsequent up- or down-conditioning. In Figure 2C, H-reflex size in neurologically unimpaired participants is plotted as a function of session number across 6 baseline sessions, 24 conditioning sessions, and 4 follow-up sessions. Data were collected from 15 participants (8 male, 7 female) of whom 2 participated in both up- and down-conditioning arms. Participants were aged 21-55 years. All the participants gave informed consent. The protocol was approved by the institutional review board (IRB) of the New York State Department of Health (approval number 05-058). Thompson et al.16 provides further details. Figure 2D shows the beneficial effect of soleus H-reflex down-conditioning in participants with chronic lower limb impairment following incomplete spinal cord injury. Successful conditioning was associated with an improvement in gait symmetry and in walking speed relative to baseline. Data were collected from 13 participants (9 male, 4 female) aged 28-68 years, who gave informed consent. The protocol was approved by the IRB of Helen Hayes Hospital (approval number 07-07). Thompson et al.14 provides further details.
Figure 2: Representative results. (A) Screenshot of the recruitment curve analysis window. (B) Screenshot of the Control Trials or Training Trials analysis window. (C) Contrasting effects of up- and down-conditioning of the soleus H-reflex in uninjured participants. Red upward triangles show mean H-reflex size from N = 6 successfully up-conditioned participants (out of 8); blue downward triangles show mean responses from N = 8 successfully down-conditioned participants (out of 9). Error bars denote standard error. This image has been modified from Thompson et al.16. (D) Therapeutic effect of soleus H-reflex down-conditioning on walking speed and gait symmetry in people with chronic impairment following incomplete spinal cord injury. The bars contrast results for N = 6 participants whose H-reflexes were successfully down-conditioned against N = 4 participants from the control condition (no operant conditioning) and N = 3 participants in whom the down-conditioning protocol failed to reduce reflex size. Error bars denote standard error. Each asterisk indicates a p-value below 0.05 on a paired t-test comparing pre- against post-conditioning measurements. This image has been modified from Thompson et al.14. Please click here to view a larger version of this figure.
The protocol described above is suitable for demonstrating soleus H-reflex down-conditioning in a typical adult without neurological impairment. The precise parameter values may vary from person to person and particularly as a function of impairment. Whereas the participant's recruitment curve reached Mmax at a stimulating current of around 25 mA in the video, another person might require 50 mA or more, so the current would be increased in larger steps during recruitment curve measurement. They might also require a longer pulse duration. A third person might be more sensitive and require smaller current settings. The protocol also needs to be adapted according to the muscle that is being conditioned. For example, when targeting the flexor carpi radialis muscle24,25, a lower current setting is generally used; the Voluntary Contraction mode should be used to establish a scale for the background-EMG limits; and greater care must be taken both during the optimization of electrode placement and during the optimization of posture, which must then be kept constant across trials.
The protocol is sensitive to variations in the relationship between stimulator current setting and the amount of current actually delivered to the nerve—this may be affected by small variations in posture, hydration of the participant, and drying out of the adhesive electrode gel. In H-reflex conditioning, this problem can be mitigated by using M-wave size as an indicator of effective stimulation intensity. It reflects the number of soleus motoneuron efferent axons excited by the stimulus. Thus, if M-wave size is kept constant, it implies that the number of primary afferent axons excited by the stimulus, i.e., the axons that elicit the H-reflex, is also kept constant (see also Crone et al.26). Hence, this M-wave is referred to as the reference response in the software. For this reason, step 4.5.12. mentions that the target M-wave size should be recorded. It is actually more important to keep this response size roughly constant than to keep the nominal current strictly constant. The Sequence tab of the analysis window allows retrospective verification of M-wave constancy over each run; for soleus H-reflex conditioning, this is often sufficient to correct any problems. For greater control, a second monitor may be attached to the computer to display real-time M-wave analytics that guide trial-by-trial manual adjustment. Automation of this control task is an ongoing project27.
Diurnal variation may also affect a person's electrophysiological responses28,29,30,31. For this reason, it is recommended that all sessions be performed at the same time of day, i.e., within the same 3 h time window.
The success of operant conditioning may be sensitive to the accuracy of the time interval chosen by the operator to define the H-reflex; in particular, the interval should not be too wide. Detailed guidelines for correct interval definition are beyond the scope of the current article. This is also a function that will be automated in future versions of the software.
A critical step in the protocol is step 4.5.6., in which the operator manually increases the stimulator current repeatedly after each fixed number of trials. Mis-counting the trials here or mis-adjusting the current dial can lead to distortion of the resulting recruitment curve. This possibility of user error can be mitigated by enabling the Digitimer Link option, which allows automation of the current adjustment for one particular stimulator model.
This article has focused on H-reflex conditioning, as it is the most fully developed of the potential clinical applications of EPOCS. The existing software helps researchers in the ongoing efforts to hone this protocol toward wide clinical dissemination32. Beyond H-reflex conditioning, EPOCS may also be applied in its current form to a wider variety of stimulation methods and evoked responses. For example, it can equally well trigger a mechanical device that elicits a stretch reflex, which may also be conditioned33,34,35. The approach is adaptable to an individual's impairments; in one person, down-conditioning the soleus H-reflex improves locomotion by reducing spastic hyperreflexia14; in another, up-conditioning the tibialis anterior MEP improves locomotion by alleviating foot drop36.
While efforts are ongoing to produce a commercial implementation of the protocol, the original software will be maintained in parallel as a research tool to provide the necessary flexibility to expand the field of targeted neuroplasticity. This flexibility is enabled by the modularity and extensibility of the widespread and well-established BCI2000 software platform, on which EPOCS is based. This means that, with minimal intervention by a software engineer, the system is re-configurable for an even wider variety of research purposes. For example, it can be configured to record additional biosignal channels or additional sensors for later analysis (e.g., foot switches and motion tracking sensors) for conditioning during locomotion. It can also be programmed to consider additional triggering criteria for stimulation (e.g., triggering stimulation only at a particular part of the gait cycle) or to trigger additional reinforcement stimuli on successful or unsuccessful trials. Example customization files are provided.
Targeted neuroplasticity is still in its infancy. Its as-yet unexplored avenues are expected to provide great benefits both for developing novel therapeutic approaches (as discussed above) and for elucidating the natural history of disease and the mechanisms of central nervous system function in both health and disease2,32,37. We are, therefore, committed to maintaining and supporting EPOCS as a key tool for realizing this therapeutic and scientific potential.
The authors have nothing to disclose.
This work was supported by NIH (NIBIB) P41EB018783 (JRW), NIH (NINDS) R01NS114279 (AKT), NIH (NINDS) U44NS114420 (I. Clements, AKT, JRW), NYS SCIRB C33279GG & C32236GG (JRW), NIH (NICHD) P2C HD086844 (S. Kautz), The Doscher Neurorehabilitation Research Program (AKT), and Stratton Albany VA Medical Center.
Alcohol swabs | any | For application to skin | |
BNC cable (long) x 1 | any | Male BNC to male BNC, long enough to reach from digitizer to stimulator | |
BNC cable (medium) x 2 | any | Male BNC to male BNC, long enough to reach from amplifier to digitizer | |
BNC cable (short) x 1 | any | Male BNC to male BNC, short (to patch between two digitizer ports) | |
BNC tee connector | any | Female-male-female BNC splitter | |
Computer | Lenovo | ThinkStation P340 | A wide range of computing hardware is suitable, especially if using a USB digitizer (no PCI slots needed). Must run Windows 7+. Include standard keyboard & mouse. |
Constant-current stimulator | Digitimer Ltd. | DS8R | The DS8R enjoys EPOCS automation support. If controlled manually, other constant-current stimulators may be used provided they have an external TTL trigger and can achieve a pulse duration of 1 ms or more. |
Digitizer (option A) | National Instruments | USB-6212 | USB digitizer with integrated BNC connectors. |
Digitizer (option B) | National Instruments | PCIe-6321 | PCIe digitizer—requires desktop computer with a free PCI slot, also cable and BNC terminal block (below) |
Digitizer cable (for option B only) | National Instruments | SHC68-68-EPM | Connects PCIe digitizer to BNC terminal block |
Digitizer terminal block (for option B only) | National Instruments | BNC-2090A | 19-inch-rack-mountable BNC terminal block |
EMG amplifier system | Bortec Biomedical Ltd. | AMT-8 | Analog amplifier + portable unit + long transmission cable + battery pack + two 500-gain active electrode leads (1 bipolar, 1 bipolar with ground) |
Monitor | any | Large enough for the participant to see clearly from the intended viewing distance. | |
NeuroPlus electrodes (22 x 22 mm) x 6 | Vermont Medical Inc. | A10040-60 | Disposable self-adhesive silver/silver-chloride 22 x 22 mm surface-EMG electrodes. 6 needed per session (11 on participant's first session) |
NeuroPlus electrode (22 x 35 mm) x 1 | Vermont Medical Inc. | A10041-60 | Disposable self-adhesive silver/silver-chloride 22 x 35 mm surface-EMG electrode. 1 needed per session. |
Snap lead x 2 | any | EDR1220 | Leads for stimulating electrodes: 1.5mm DIN to button snap |
Wire | any | 8–10 cm length of single-core insulated wire |