Operant conditioning of specific patterns of neural and muscular activity

Computers and neurosurgery

At the turn of the twentieth century, the only computational device used in neurosurgical procedures was the brain of the surgeon. Today, most neurosurgical procedures rely at least in part on the use of a computer to help perform surgeries accurately and safely. The techniques that revolutionized neurosurgery were mostly developed after the 1950s. Just before that era, the transistor was invented in the late 1940s, and the integrated circuit was invented in the late 1950s. During this time, the first automated, programmable computational machines were introduced. The rapid progress in the field of neurosurgery not only occurred hand in hand with the development of modern computers, but one also can state that modern neurosurgery would not exist without computers. The focus of this article is the impact modern computers have had on the practice of neurosurgery. Neuroimaging, neuronavigation, and neuromodulation are examples of tools in the armamentarium of the modern neurosurgeon that owe each step in their evolution to progress made in computer technology. Advances in computer technology central to innovations in these fields are highlighted, with particular attention to neuroimaging. Developments over the last 10 years in areas of sensors and robotics that promise to transform the practice of neurosurgery further are discussed. Potential impacts of advances in computers related to neurosurgery in developing countries and underserved regions are also discussed. As this article illustrates, the computer, with its underlying and related technologies, is central to advances in neurosurgery over the last half century.

Dissociating motor cortex from the motor

During closed-loop control of a brain-computer interface, neurons in the primary motor cortex can be intensely active even though the subject may be making no detectable movement or muscle contraction. How can neural activity in the primary motor cortex become dissociated from the movements and muscles of the native limb that it normally controls? Here we examine circumstances in which motor cortex activity is known to dissociate from movement--including mental imagery, visuo-motor dissociation and instructed delay. Many such motor cortex neurons may be related to muscle activity only indirectly. Furthermore, the integration of thousands of synaptic inputs by individual α-motoneurons means that under certain circumstances even cortico-motoneuronal cells, which make monosynaptic connections to α-motoneurons, can become dissociated from muscle activity. The natural ability of motor cortex neurons under voluntarily control to become dissociated from bodily movement may underlie the utility of this cortical area for controlling brain-computer interfaces.

Interfacing with the computational brain

Neuroscience is just beginning to understand the neural computations that underlie our remarkable capacity to learn new motor tasks. Studies of natural movements have emphasized the importance of concepts such as dimensionality reduction within hierarchical levels of redundancy, optimization of behavior in the presence of sensorimotor noise and internal models for predictive control. These concepts also provide a framework for understanding the improvements in performance seen in myoelectric-controlled interface and brain-machine interface paradigms. Recent experiments reveal how volitional activity in the motor system combines with sensory feedback to shape neural representations and drives adaptation of behavior. By elucidating these mechanisms, a new generation of intelligent interfaces can be designed to exploit neural plasticity and restore function after neurological injury.

The Asilomar Survey: Stakeholders' Opinions on Ethical Issues Related to Brain-Computer Interfacing

Brain-Computer Interface (BCI) research and (future) applications raise important ethical issues that need to be addressed to promote societal acceptance and adequate policies. Here we report on a survey we conducted among 145 BCI researchers at the 4^th International BCI conference, which took place in May-June 2010 in Asilomar, California. We assessed respondents' opinions about a number of topics. First, we investigated preferences for terminology and definitions relating to BCIs. Second, we assessed respondents' expectations on the marketability of different BCI applications (BCIs for healthy people, BCIs for assistive technology, BCIs-controlled neuroprostheses and BCIs as therapy tools). Third, we investigated opinions about ethical issues related to BCI research for the development of assistive technology: informed consent process with locked-in patients, risk-benefit analyses, team responsibility, consequences of BCI on patients' and families' lives, liability and personal identity and interaction with the media. Finally, we asked respondents which issues are urgent in BCI research.

Volitional control of single cortical neurons in a brain-machine interface

Volitional control of cortical activity is relevant for optimizing control signals for neuroprosthetic devices. We explored the control of firing rates of single cortical cells in two Macaca nemestrina monkeys by providing visual feedback of neural activity and rewarding changes in cell rates. During 'brain-control' sessions, the monkeys modulated the activity of each of 246 cells to acquire targets requiring high or low discharge rates. Cell control improved more than two-fold from the beginning of practice to peak performance. Cell activity was modulated substantially more during brain control than during wrist movements. When recording stability permitted, the monkeys practiced controlling activity of the same cells across multiple days. The performance improved substantially for 27 of 36 cells when practicing brain control across days. The monkeys maintained discharge rates within each target for 1 s, but could maintain rates for up to 3 s for some cells, and performed the brain-control task equally well using cells recorded from the pre-central cortex compared to cells in the post-central cortex, and independently of any directional tuning. These findings demonstrate that arbitrary single cortical neurons, regardless of the strength of directional tuning, are capable of controlling cursor movements in a one-dimensional brain-machine interface. It is possible that direct conversion of activity from single cortical cells to control signals for neuroprosthetic devices may be a useful complementary strategy to population decoding of the intended movement direction.

Learning the microstructure of successful behavior

Reinforcement signals indicating success or failure are known to alter the probability of selecting between distinct actions. However, successful performance of many motor skills, such as speech articulation, also requires learning behavioral trajectories that vary continuously over time. Here, we investigated how temporally discrete reinforcement signals shape a continuous behavioral trajectory, the fundamental frequency of adult Bengalese finch song. We provided reinforcement contingent on fundamental frequency performance only at one point in the song. Learned changes to fundamental frequency were maximal at this point, but also extended both earlier and later in the fundamental frequency trajectory. A simple principle predicted the detailed structure of learning: birds learned to produce the average of the behavioral trajectories associated with successful outcomes. This learning rule accurately predicted the structure of learning at a millisecond timescale, demonstrating that the nervous system records fine-grained details of successful behavior and uses this information to guide learning.

Inference from populations: going beyond models

How are abstract signals, like intent, represented in neural populations? By creating a direct link between neural activity and behavior, brain-computer interfaces (BCIs) can help answer this question. Early instantiations of these devices sought mainly to mimic arm movements: by building models of arm tuning for the neurons, desired arm movements could be read out and used to control various prosthetic devices. However, as the functionality of these devices increases, a more general approach that relies less on endogenous control signals may be required. Here we review some of the current, model-based approaches for finding volitional control signals for spiking-based BCIs, and present some new approaches for finding control signals without resorting to parametric models of neural activity.

Evolution of brain-computer interface: action potentials, local field potentials and electrocorticograms

Brain computer interfaces (BCIs) were originally developed to give severely motor impaired patients a method to communicate and interact with their environment. Initially most BCI systems were based on non-invasive electroencephalographic recordings from the surface of the scalp. To increase control speed, accuracy and complexity, researchers began utilizing invasive recording modalities. BCIs using multi-single unit action potentials have provided elegant multi-dimensional control of both computer cursors and robotic limbs in the last few years. However, long-term stability issues with single-unit arrays has lead researchers to investigate other invasive recording modalities such as high-frequency local field potentials and electrocorticography (ECoG). Although ECoG originally evolved as a replacement for single-unit BCIs, it has come full circle to become an effective tool for studying cortical neurophysiology.

Human cortical prostheses: lost in translation?

Direct brain control of a prosthetic system is the subject of much popular and scientific news. Neural technology and science have advanced to the point that proof-of-concept systems exist for cortically-controlled prostheses in rats, monkeys, and even humans. However, realizing the dream of making such technology available to everyone is still far off. Fortunately today there is great public and scientific interest in making this happen, but it will only occur when the functional benefits of such systems outweigh the risks. In this article, the authors briefly summarize the state of the art and then highlight many issues that will directly limit clinical translation, including system durability, system performance, and patient risk. Despite the challenges, scientists and clinicians are in the desirable position of having both public and fiscal support to begin addressing these issues directly. The ultimate challenge now is to determine definitively whether these prosthetic systems will become clinical reality or forever unrealized.

Principles of neural ensemble physiology underlying the operation of brain-machine interfaces

Research on brain-machine interfaces has been ongoing for at least a decade. During this period, simultaneous recordings of the extracellular electrical activity of hundreds of individual neurons have been used for direct, real-time control of various artificial devices. Brain-machine interfaces have also added greatly to our knowledge of the fundamental physiological principles governing the operation of large neural ensembles. Further understanding of these principles is likely to have a key role in the future development of neuroprosthetics for restoring mobility in severely paralysed patients.

Applications of real-time fMRI

For centuries people have aspired to understand and control the functions of the mind and brain. It has now become possible to image the functioning of the human brain in real time using functional MRI (fMRI), and thereby to access both sides of the mind-brain interface--subjective experience (that is, one's mind) and objective observations (that is, external, quantitative measurements of one's brain activity)--simultaneously. Developments in neuroimaging are now being translated into many new potential practical applications, including the reading of brain states, brain-computer interfaces, communicating with locked-in patients, lie detection, and learning control over brain activation to modulate cognition or even treat disease.

Brain-computer symbiosis

The theoretical groundwork of the 1930s and 1940s and the technical advance of computers in the following decades provided the basis for dramatic increases in human efficiency. While computers continue to evolve, and we can still expect increasing benefits from their use, the interface between humans and computers has begun to present a serious impediment to full realization of the potential payoff. This paper is about the theoretical and practical possibility that direct communication between the brain and the computer can be used to overcome this impediment by improving or augmenting conventional forms of human communication. It is about the opportunity that the limitations of our body's input and output capacities can be overcome using direct interaction with the brain, and it discusses the assumptions, possible limitations and implications of a technology that I anticipate will be a major source of pervasive changes in the coming decades.

Rapid changes in throughput from single motor cortex neurons to muscle activity

Motor cortex output is capable of considerable reorganization, which involves modulation of excitability within the cortex. Does such reorganization also involve changes beyond the cortex, at the level of throughput from single motor cortex neurons to muscle activity? We examined such throughput during a paradigm that provided incentive for enhancing functional connectivity from motor cortex neurons to muscles. Short-latency throughput from a recorded neuron to muscle activity not present during some behavioral epochs often appeared during others. Such changes in throughput could not always be attributed to a higher neuron firing rate, to more ongoing muscle activity, or to neuronal synchronization, indicating that reorganization of motor cortex output may involve rapid changes in functional connectivity from single motor cortex neurons to alpha-motoneuron pools.

The development of brain-machine interface neuroprosthetic devices

The development of brain-machine interface technology is a logical next step in the overall direction of neuroprosthetics. Many of the required technological advances that will be required for clinical translation of brain-machine interfaces are already under development, including a new generation of recording electrodes, the decoding and interpretation of signals underlying intention and planning, actuators for implementation of mental plans in virtual or real contexts, direct somatosensory feedback to the nervous system to refine actions, and training to encourage plasticity in neural circuits. Although pre-clinical studies in nonhuman primates demonstrate high efficacy in a realistic motor task with motor cortical recordings, there are many challenges in the clinical translation of even simple tasks and devices. Foremost among these challenges is the development of biocompatible electrodes capable of long-term, stable recording of brain activity and implantable amplifiers and signal processors that are sufficiently resistant to noise and artifact to faithfully transmit recorded signals to the external environment. Whether there is a suitable market for such new technology depends on its efficacy in restoring and enhancing neural function, its risks of implantation, and its long-term efficacy and usefulness. Now is a critical time in brain-machine interface development because most ongoing studies are science-based and noncommercial, allowing new approaches to be included in commercial schemes under development.

Reading and controlling human brain activation using real-time functional magnetic resonance imaging

Understanding how to control how the brain's functioning mediates mental experience and the brain's processing to alter cognition or disease are central projects of cognitive and neural science. The advent of real-time functional magnetic resonance imaging (rtfMRI) now makes it possible to observe the biology of one's own brain while thinking, feeling and acting. Recent evidence suggests that people can learn to control brain activation in localized regions, with corresponding changes in their mental operations, by observing information from their brain while inside an MRI scanner. For example, subjects can learn to deliberately control activation in brain regions involved in pain processing with corresponding changes in experienced pain. This may provide a novel, non-invasive means of observing and controlling brain function, potentially altering cognitive processes or disease.

Comparing effects in spike-triggered averages of rectified EMG across different behaviors

Effects in spike-triggered averages (SpikeTAs) of rectified electromyographic activity (EMG) compiled for the same neuron-muscle pair during various behaviors often appear different. Do these differences represent significant changes in the effect of the neuron on the muscle activity? Quantitative comparison of such differences has been limited by two methodological problems, which we address here. First, although the linear baseline trend of many SpikeTAs can be adjusted with ramp subtraction, the curvilinear baseline trend of other SpikeTAs can not. To address this problem, we estimated baseline trends using a form of moving average. Artificial triggers were created in 1 ms increments from 40 ms before to 40 ms after each spike used to compile the SpikeTA. These 81 triggers were used to compile another average of rectified EMG, which we call a single-spike increment-shifted average (single-spike ISA). Single-spike ISAs were averaged to produce an overall ISA, which captured slow trends in the baseline EMG while distributing any spike-locked features evenly throughout the 80 ms analysis window. The overall ISA then was subtracted from the initial SpikeTA, removing any slow baseline trends for more accurate measurement of SpikeTA effects. Second, the measured amplitude and temporal characteristics of SpikeTA effects produced by the same neuron-muscle pair may vary during different behaviors. But whether or not such variation is significant has been difficult to ascertain. We therefore applied a multiple fragment approach to permit statistical comparison of the measured features of SpikeTA effects for the same neuron-muscle pair during different behavioral epochs. Spike trains recorded in each task were divided into non-overlapping fragments of 100 spikes each, and a separate, ISA-corrected, SpikeTA was compiled for each fragment. Measurements made on these fragment SpikeTAs then were used as test statistics for comparison of peak percent increase, mean percent increase, peak width at half maximum, onset latency, and offset latency. The average of each test statistic measured from the fragment SpikeTAs was well correlated with the single measurement made on the overall SpikeTA. The multiple fragment approach provides a sensitive means of identifying significant changes in SpikeTA effects.

Neural events in the reinforcement contingency

When neural events are analyzed as stimuli and responses, functional relations among them and among overt stimuli and responses can be unveiled. The integration of neuroscience and the experimental analysis of behavior is beginning to provide empirical evidence of involvement of neural events in the three-term contingency relating discriminative stimuli, responses, and consequences. This paper is aimed at highlighting exemplar instances in the development of this issue. It has long been known that the electrical stimulation of certain cerebral areas can have a reinforcing function. Extraordinary technological advances in recent years show that neural activity can be selected by consequences. For example, the activity of in vitro isolated neurons that receive dopamine as a reinforcer functions as a cellular analogue of operant conditioning. The in vivo activity of populations of neurons of rats and monkeys can be recorded on an instant-to-instant basis and can then be used to move mechanical arms or track a target as a function of consequences. Neural stimulation acts as a discriminative stimulus for operant responses that are in turn maintained by neural consequences. Together with investigations on the molecular basis of classical conditioning, those studies are examples of possibilities that are being created for the study of behavior-environment interactions within the organism. More important, they show that, as an element in the three-term contingency, neural activity follows the same laws as other events.

Brain–machine interfaces: past, present and future

Since the original demonstration that electrical activity generated by ensembles of cortical neurons can be employed directly to control a robotic manipulator, research on brain-machine interfaces (BMIs) has experienced an impressive growth. Today BMIs designed for both experimental and clinical studies can translate raw neuronal signals into motor commands that reproduce arm reaching and hand grasping movements in artificial actuators. Clearly, these developments hold promise for the restoration of limb mobility in paralyzed subjects. However, as we review here, before this goal can be reached several bottlenecks have to be passed. These include designing a fully implantable biocompatible recording device, further developing real-time computational algorithms, introducing a method for providing the brain with sensory feedback from the actuators, and designing and building artificial prostheses that can be controlled directly by brain-derived signals. By reaching these milestones, future BMIs will be able to drive and control revolutionary prostheses that feel and act like the human arm.

Neuronal ensemble control of prosthetic devices by a human with tetraplegia

Neuromotor prostheses (NMPs) aim to replace or restore lost motor functions in paralysed humans by routeing movement-related signals from the brain, around damaged parts of the nervous system, to external effectors. To translate preclinical results from intact animals to a clinically useful NMP, movement signals must persist in cortex after spinal cord injury and be engaged by movement intent when sensory inputs and limb movement are long absent. Furthermore, NMPs would require that intention-driven neuronal activity be converted into a control signal that enables useful tasks. Here we show initial results for a tetraplegic human (MN) using a pilot NMP. Neuronal ensemble activity recorded through a 96-microelectrode array implanted in primary motor cortex demonstrated that intended hand motion modulates cortical spiking patterns three years after spinal cord injury. Decoders were created, providing a 'neural cursor' with which MN opened simulated e-mail and operated devices such as a television, even while conversing. Furthermore, MN used neural control to open and close a prosthetic hand, and perform rudimentary actions with a multi-jointed robotic arm. These early results suggest that NMPs based upon intracortical neuronal ensemble spiking activity could provide a valuable new neurotechnology to restore independence for humans with paralysis.

The emerging world of motor neuroprosthetics: a neurosurgical perspective

A MOTOR NEUROPROSTHETIC device, or brain computer interface, is a machine that can take some type of signal from the brain and convert that information into overt device control such that it reflects the intentions of the user's brain. In essence, these constructs can decode the electrophysiological signals representing motor intent. With the parallel evolution of neuroscience, engineering, and rapid computing, the era of clinical neuroprosthetics is approaching as a practical reality for people with severe motor impairment. Patients with such diseases as spinal cord injury, stroke, limb loss, and neuromuscular disorders may benefit through the implantation of these brain computer interfaces that serve to augment their ability to communicate and interact with their environment. In the upcoming years, it will be important for the neurosurgeon to understand what a brain computer interface is, its fundamental principle of operation, and what the salient surgical issues are when considering implantation. We review the current state of the field of motor neuroprosthetics research, the early clinical applications, and the essential considerations from a neurosurgical perspective for the future.

Changes in visual receptive fields with microstimulation of frontal cortex

The influence of attention on visual cortical neurons has been described in terms of its effect on the structure of receptive fields (RFs), where multiple stimuli compete to drive neural responses and ultimately behavior. We stimulated the frontal eye field (FEF) of passively fixating monkeys and produced changes in V4 responses similar to known effects of voluntary attention. Subthreshold FEF stimulation enhanced visual responses at particular locations within the RF and altered the interaction between pairs of RF stimuli to favor those aligned with the activated FEF site. Thus, we could influence which stimulus drove the responses of individual V4 neurons. These results suggest that spatial signals involved in saccade preparation are used to covertly select among multiple stimuli appearing within the RFs of visual cortical neurons.

Selection and parameterization of cortical neurons for neuroprosthetic control

When designing neuroprosthetic interfaces for motor function, it is crucial to have a system that can extract reliable information from available neural signals and produce an output suitable for real life applications. Systems designed to date have relied on establishing a relationship between neural discharge patterns in motor cortical areas and limb movement, an approach not suitable for patients who require such implants but who are unable to provide proper motor behavior to initially tune the system. We describe here a method that allows rapid tuning of a population vector-based system for neural control without arm movements. We trained highly motivated primates to observe a 3D center-out task as the computer played it very slowly. Based on only 10-12 s of neuronal activity observed in M1 and PMd, we generated an initial mapping between neural activity and device motion that the animal could successfully use for neuroprosthetic control. Subsequent tunings of the parameters led to improvements in control, but the initial selection of neurons and estimated preferred direction for those cells remained stable throughout the remainder of the day. Using this system, we have observed that the contribution of individual neurons to the overall control of the system is very heterogeneous. We thus derived a novel measure of unit quality and an indexing scheme that allowed us to rate each neuron's contribution to the overall control. In offline tests, we found that fewer than half of the units made positive contributions to the performance. We tested this experimentally by having the animals control the neuroprosthetic system using only the 20 best neurons. We found that performance in this case was better than when the entire set of available neurons was used. Based on these results, we believe that, with careful task design, it is feasible to parameterize control systems without any overt behaviors and that subsequent control system design will be enhanced with cautious unit selection. These improvements can lead to systems demanding lower bandwidth and computational power, and will pave the way for more feasible clinical systems.

Naive coadaptive cortical control

The ability to control a prosthetic device directly from the neocortex has been demonstrated in rats, monkeys and humans. Here we investigate whether neural control can be accomplished in situations where (1) subjects have not received prior motor training to control the device (naive user) and (2) the neural encoding of movement parameters in the cortex is unknown to the prosthetic device (naive controller). By adopting a decoding strategy that identifies and focuses on units whose firing rate properties are best suited for control, we show that naive subjects mutually adapt to learn control of a neural prosthetic system. Six untrained Long-Evans rats, implanted with silicon micro-electrodes in the motor cortex, learned cortical control of an auditory device without prior motor characterization of the recorded neural ensemble. Single- and multi-unit activities were decoded using a Kalman filter to represent an audio "cursor" (90 ms tone pips ranging from 250 Hz to 16 kHz) which subjects controlled to match a given target frequency. After each trial, a novel adaptive algorithm trained the decoding filter based on correlations of the firing patterns with expected cursor movement. Each behavioral session consisted of 100 trials and began with randomized decoding weights. Within 7 +/- 1.4 (mean +/- SD) sessions, all subjects were able to significantly score above chance (P < 0.05, randomization method) in a fixed target paradigm. Training lasted 24 sessions in which both the behavioral performance and signal to noise ratio of the peri-event histograms increased significantly (P < 0.01, ANOVA). Two rats continued training on a more complex task using a bilateral, two-target control paradigm. Both subjects were able to significantly discriminate the target tones (P < 0.05, Z-test), while one subject demonstrated control above chance (P < 0.05, Z-test) after 12 sessions and continued improvement with many sessions achieving over 90% correct targets. Dynamic analysis of binary trial responses indicated that early learning for this subject occurred during session 6. This study demonstrates that subjects can learn to generate neural control signals that are well suited for use with external devices without prior experience or training.

Optimizing a linear algorithm for real-time robotic control using chronic cortical ensemble recordings in monkeys

Previous work in our laboratory has demonstrated that a simple linear model can be used to translate cortical neuronal activity into real-time motor control commands that allow a robot arm to mimic the intended hand movements of trained primates. Here, we describe the results of a comprehensive analysis of the contribution of single cortical neurons to this linear model. Key to the operation of this model was the observation that a large percentage of cortical neurons located in both frontal and parietal cortical areas are tuned for hand position. In most neurons, hand position tuning was time-dependent, varying continuously during a 1-sec period before hand movement onset. The relevance of this physiological finding was demonstrated by showing that maximum contribution of individual neurons to the linear model was only achieved when optimal parameters for the impulse response functions describing time-varying neuronal position tuning were selected. Optimal parameters included impulse response functions with 1.0- to 1.4-sec time length and 50- to 100-msec bins. Although reliable generalization and long-term predictions (60-90 min) could be achieved after 10-min training sessions, we noticed that the model performance degraded over long periods. Part of this degradation was accounted by the observation that neuronal position tuning varied significantly throughout the duration (60-90 min) of a recording session. Altogether, these results indicate that the experimental paradigm described here may be useful not only to investigate aspects of neural population coding, but it may also provide a test bed for the development of clinically useful cortical prosthetic devices aimed at restoring motor functions in severely paralyzed patients.

A brain-computer interface using electrocorticographic signals in humans

Brain-computer interfaces (BCIs) enable users to control devices with electroencephalographic (EEG) activity from the scalp or with single-neuron activity from within the brain. Both methods have disadvantages: EEG has limited resolution and requires extensive training, while single-neuron recording entails significant clinical risks and has limited stability. We demonstrate here for the first time that electrocorticographic (ECoG) activity recorded from the surface of the brain can enable users to control a one-dimensional computer cursor rapidly and accurately. We first identified ECoG signals that were associated with different types of motor and speech imagery. Over brief training periods of 3-24 min, four patients then used these signals to master closed-loop control and to achieve success rates of 74-100% in a one-dimensional binary task. In additional open-loop experiments, we found that ECoG signals at frequencies up to 180 Hz encoded substantial information about the direction of two-dimensional joystick movements. Our results suggest that an ECoG-based BCI could provide for people with severe motor disabilities a non-muscular communication and control option that is more powerful than EEG-based BCIs and is potentially more stable and less traumatic than BCIs that use electrodes penetrating the brain.

Learning to control a brain-machine interface for reaching and grasping by primates

Reaching and grasping in primates depend on the coordination of neural activity in large frontoparietal ensembles. Here we demonstrate that primates can learn to reach and grasp virtual objects by controlling a robot arm through a closed-loop brain–machine interface (BMIc) that uses multiple mathematical models to extract several motor parameters (i.e., hand position, velocity, gripping force, and the EMGs of multiple arm muscles) from the electrical activity of frontoparietal neuronal ensembles. As single neurons typically contribute to the encoding of several motor parameters, we observed that high BMIc accuracy required recording from large neuronal ensembles. Continuous BMIc operation by monkeys led to significant improvements in both model predictions and behavioral performance. Using visual feedback, monkeys succeeded in producing robot reach-and-grasp movements even when their arms did not move. Learning to operate the BMIc was paralleled by functional reorganization in multiple cortical areas, suggesting that the dynamic properties of the BMIc were incorporated into motor and sensory cortical representations.

With visual feedback, macaque monkeys learn to control a robot arm through a neural interface which records activity from multiple cortical areas

Modulation of electroencephalographic responses to transcranial magnetic stimulation: evidence for changes in cortical excitability related to movement

Transcranial magnetic stimulation (TMS) and multichannel electroencephalography (EEG) were used for the investigation of cortical excitability preceding voluntary movement in human subjects. The study showed the practical value of the combined TMS-EEG approach in differentiating between cortical and spinal-cord mechanisms, which is difficult with conventional electromyographic measures alone. TMS induced a pronounced negativity (N100) lasting for 150-200 ms, with the amplitude maximum in the stimulated hemisphere. When TMS was applied just before the onset of the visually triggered movement, N100 was markedly attenuated, although motor evoked potentials (MEPs) became larger. We suggest that the N100 component represents an inhibitory response following TMS. This interpretation is in agreement with intracellular recordings in animals, paired-pulse TMS studies and experiments showing increased premovement excitability on the basis of MEPs. N100 was not affected only by the subsequent movement, but also by the switching from rest to the motor-task condition, which caused a slight attenuation of the N100 component; no changes, however, were found in the amplitude of MEPs, suggesting that modified excitability did not affect the output of the corticospinal pyramidal cells. By contrast to MEPs, N100 was modulated also by the presentation of the visual stimulus alone, i.e. when no movement was required. This attenuation suggests that even in a rest condition visual stimuli have an access to the sensorimotor regions of the cortex, most probably through ascending arousal brain systems.

The time course of changes in motor cortex excitability associated with voluntary movement

The excitability of the motor cortex is modulated before and after voluntary movements. Transcranial magnetic stimulation studies showed increased corticospinal excitability from about 80 and 100 ms before EMG onset for simple reaction time and self-paced movements, respectively. Following voluntary movements, there are two phases of increased corticospinal excitability from 0 to approximately 100 ms and from approximately 100 to 160 ms after EMG offset. The first phase may correspond to the frontal peak of motor potential in movement-related cortical potentials studies and the movement-evoked magnetic field I (MEFI) in magnetoencephalographic (MEG) studies, and likely represents a time when decreasing output from the motor cortex falls below that required for activation of spinal motoneurons, but is still above resting levels. The second phase of increased corticospinal excitability may be due to peripheral proprioceptive inputs or may be centrally programmed representing a subthreshold, second agonist burst. This may correspond to the MEFII in MEG studies. Corticospinal excitability was reduced below baseline levels from about 500 to 1,000 ms after EMG offset, similar to the timing of increase in the power (event-related synchronization, ERS) of motor cortical rhythm. Similarly, motor cortex excitability is reduced at the time of ERS of motor cortical rhythm following median nerve stimulation. These findings support the hypothesis that ERS represents an inactive, idling state of the cortex. The time course of cortical activation is abnormal in movement disorders such as Parkinson's disease and dystonia, reflecting abnormalities in both movement preparation and in cortical excitability following movement.

Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex

To determine whether simultaneously recorded motor cortex neurons can be used for real-time device control, rats were trained to position a robot arm to obtain water by pressing a lever. Mathematical transformations, including neural networks, converted multineuron signals into 'neuronal population functions' that accurately predicted lever trajectory. Next, these functions were electronically converted into real-time signals for robot arm control. After switching to this 'neurorobotic' mode, 4 of 6 animals (those with > 25 task-related neurons) routinely used these brain-derived signals to position the robot arm and obtain water. With continued training in neurorobotic mode, the animals' lever movement diminished or stopped. These results suggest a possible means for movement restoration in paralysis patients.

In vitro analog of operant conditioning in aplysia. II. Modifications of the functional dynamics of an identified neuron contribute to motor pattern selection

Previously, an analog of operant conditioning was developed using the buccal ganglia of Aplysia, the probabilistic occurrences of a specific motor pattern (i.e., pattern I), a contingent reinforcement (i.e., stimulation of the esophageal nerve), and monotonic stimulation of a peripheral nerve (i.e., n.2,3). This analog expressed a key feature of operant conditioning (i.e., selective enhancement of the probability of occurrence of a designated motor pattern by contingent reinforcement). In addition, the training induced changes in the dynamical properties of neuron B51, an element of the buccal central pattern generator. To gain insights into the neuronal mechanisms that mediate features of operant conditioning, the present study identified a neuronal element that was critically involved in the selective enhancement of pattern I. We found that bursting activity in cell B51 contributed significantly to the expression of pattern I and that changes in the dynamical properties of this cell were associated with the selective enhancement of pattern I. These changes could be induced by an explicit association of reinforcement with random depolarization of B51. No stimulation of n.2,3 was required. These results indicate that the selection of a designated motor pattern by contingent reinforcement and the underlying neuronal plasticity resulted from the association of reinforcement with a component of central neuronal activity that contributes to a specific motor pattern. The sensory stimulus that allows for occurrences of different motor acts may not be critical for induction of plasticity that mediates the selection of a motor output by contingent reinforcement in operant conditioning.

Time course of corticospinal excitability in reaction time and self-paced movements

We used transcranial magnetic stimulation (TMS) to study the time course of corticospinal excitability before and after brisk thumb abduction movements, either in a simple reaction time (RT) paradigm or self-paced. Premovement increase in corticospinal excitability began about 20 msec earlier for self-paced compared with simple RT movements. For both simple RT and self-paced movements after electromyographic (EMG) offset, there was a first period of increased excitability from 0 to 100 msec, followed by a second period from 100 to 160 msec. Corticospinal excitability was decreased from about 500 to 1,000 msec after EMG offset for both types of movements. Our results show that motor preparation that begins 1.5 to 2 seconds before self-paced movement is not associated with increased corticospinal excitability. The first phase of increased corticospinal excitability after EMG offset may be due to activity of motor cortex neuron subthreshold for activating spinal motor neurons, and the second phase may reflect a subthreshold second agonist burst. The period of decreased corticospinal excitability after movement corresponds to the onset of event-related synchronization (ERS) of electroencephalographic signals in the 20-Hz band, and supports the hypothesis that ERS may be related to an inactive, idling state of the motor cortex.

Learning and rhythmic human EMG in ecological perspective

Previous evidence of strong interactions between learning and human treadmill locomotion led to a simplified system for studying learned rhythms in a framework of behavioral ecology. Motor control combined with instrumental conditioning in a rhythmic hand task with repeating trials, blocks, and complete regimens. Regimen contexts differed with respect to the pattern of stimulation before and after an electromyographic (EMG) response. Both an antecedent stimulus (a light flash) and a consequent stimulus (a tone indicating success or failure) were necessary for conditioning. Arguments were given for defining reinforcement as a composite of interdependent and size-scaled processes, some including knowledge of results, instead of as a single event after a response.

Operant conditioning in motor and neural integration

Empirical and theoretical reasons were given to investigate operant conditioning in a new, integrative approach within motor control physiology. Elements of inborn and learned behavior were presented in a framework specifying their stimuli and responses. The operant was redefined as a controlling discriminative stimulus, Sd, together with the response, R, it produces, on the basis of a previous literature of operant and instrumental research. Complex motor and neural activity were reviewed in accordance with partitioning of: responses, controlling stimulation, reinforcement, and functions of movement-produced stimulation. Schematics portrayed reinforcement principles through analysis of a fast pathway from Ia muscle spindle afferents to motor outflow. Methods were suggested to minimize operant units through selective reinforcement and establish them to defined end points of learning within composite, ongoing behavior. It was argued that operant neural mechanisms can be investigated efficiently only by starting with individual operants that are thoroughly characterized.

Operant conditioning of somatosensory evoked potential (SEP) amplitude in rats. II. Associated changes in reflex and continuous non-timelocked movements

Animals were rewarded for increasing (uptrain) or decreasing (downtrain) the amplitude of a 30 ms surface positive component of a somatosensory evoked potential (SEP) evoked by innocuous stimulation of the spinal trigeminal tract. The reflex movement produced by the evoking stimulus had a larger amplitude in uptraining than downtraining. This change in reflex amplitude suggests that operantly conditioning SEP amplitude was correlated with a change in innocuous somatosensory activity. There was no change in continuous non-timelocked movement associated with conditioning. This latter finding suggests that SEP conditioning is not necessarily mediated by such movement.

Jaw movements evoked by substantia nigra stimulation: role of extranigral current spread

Electrical stimulation of the substantia nigra in cats elicits jaw movements. A variety of tests were performed to assess the contribution of current spread to this phenomenon. The results of these tests strongly suggest that stimulation-evoked movements are not due to activation of the substantia nigra but rather are the result of current spread to adjacent structures.

Operant control of precentral neurons in monkeys: evidence against open loop control

Four normal monkeys were operantly conditioned to change the firing pattern of 111 precentral neurons from phasic to tonic using an operant paradigm which quantifies the control of single neurons. Two monkeys then had their contralateral pyramidal tract (PT) sectioned and one monkey had C5-7 ventral rhizotomies. Postlesion data were: (1) contralateral C1-2PT lesions did not encumber the monkeys' control of precentral PTNs: (2) contralateral C5-7 ventral rhizotomies completely abolished accurate control of precentral neurons which received proprioceptive feedback from flaccid arm regions. These results indicate that precentral neurons are operantly controlled through proprioceptive feedback from peripheral mechanoreceptors. The output of the mechanoreceptors is probably dependent upon discrete joint angles and/or muscle tension which is maintained through non-PT pathways. These data do not support the concept that precentral neurons are operantly controlled directly from a central; 'open loop', pathway.

Instrumental cardiovascular conditioning: a review

This paper reviews experiments, hypotheses, and current controversies about instrumental cardiovascular conditioning. Demonstrations of such conditioning in curarized animals challenged a differentiation between instrumental and classical learning on the basis of their respective effector systems but did not prove direct operant learning by the autonomic nervous system. In humans, ethical prohibition of curarization and lack of adequate controls for respiration and muscle tension have resulted in incomplete understanding of the roles of voluntary, somatic mediators. Despite a variety of potential clinical applications of biofeedback, the available literature lacks studies of its efficacy compared to more standard modes of therapy. The physiological mechanisms and central neural pathways involved in instrumental cardiovascular conditioning remain almost totally unknown.

Operant control of pyramidal tract neurons: the role of spinal dorsal columns

A monkey was trained to control the firing patterns of precentral pyramidal tract neurons. The operant task was for the monkey to produce consecutive interspike intervals (ISI) within a requisite range, or target. The mean time off-target (error) is used to quantify the accuracy of control the monkey could assert over each PTN. Following partial destruction of the dorsal funiculi the number of PTNs driven by peripheral stimuli greatly decreased. Those PTNs which remained responsive to peripheral stimuli were as accurately controlled as those tested before column section, whereas, those PTNs unresponsive to peripheral stimuli were significantly less accurately controlled. The conclusion is that the monkey relies heavily upon proprioceptive feedback to operantly control precentral PTNs.