Table of contents

There is a very strong interest in the feasibility of a brain machine interface as indicated by the number of submissions, citations and downloads from the Journal of Neural Engineering in this area (see the previous editorial, Durand 2006 J. Neural Eng.3 (3). This interest is clearly understandable. The ability to decode and interpret brain signals could be of critical importance for not only understanding how the brain works but also to control prosthetic devices by pure thought. However, there is a lot more to neural interfacing than the brain machine interface.

The Neural Interfaces Workshop was held at the Bethesda North Marriott Hotel and Conference Center on 21–23 August, 2006. This meeting included functional neuromuscular/electrical stimulation, auditory prosthesis, cortical prosthesis, microelectrode array technology as well as brain computer/machine interfaces. The meeting was attended by a diverse group of scientists, engineers, and clinicians, representing the basic and applied science aspects of neural interfaces. Support for this meeting came from the following institutes within the NIH: the National Institute of Neurological Disorders and Stroke (NINDS), National Institute on Aging, National Institute of Biomedical Imaging and Bioengineering, National Institute of Mental Health, and National Institute on Deafness and other Communication Disorders.

This issue of the Journal of Neural Engineering has a dedicated section covering special aspects of neural interfacing presented at the conference. A report on the presentations and the discussions at the conference written by NIH program officers can be found on page S137 Chen et al. Several papers invited by the Journal of Neural Engineering were chosen to reflect the breadth of neural interfacing within the field of neural engineering. The papers cover a wide range of neural systems from molecular interfacing to clinical restoration of standing/walking in paralyzed patients. The first paper from the Biomedical Engineering Department at Stanford University (Aravanis et al, S143; see also the front cover of the issue) is the first manuscript published in the Journal of Neural Engineering on optogenetic interfacing technology. Neurons in the central nervous system can be genetically modified to respond to light activation in order to control their activity. The second paper from the Case Western Reserve University (Tesfayesus and Durand, page S157 reports on the developments of new methods to recover nerve fascicular signals within the peripheral nervous system. The third paper by Weber et al on page S168 (University of Pittsburg and University of Alberta) describes a set of experiments involving a neural interface with dorsal root ganglia. As noted by the reviewer: `the manuscript is of high interest because it involves DRG recordings in awake behaving animals and attempts to use a combination of recordings to predict the locomotor state'. The final paper (Mushahwar et al, page S181) reviews four different interfacing technologies for restoring standing/walking in patients with spinal cord injuries: surface functional electrical stimulation, implanted muscle-based electrodes, peripheral nerve intraneural arrays and intraspinal microstimulation (ISMS) of the lumbosacral spinal cord.

The Neural Interfaces Workshop has now grown too big and will be replaced by the Neural Interfaces Conference. This change from a NIH-led workshop to a NIH-sponsored conference reflects the strong leadership from the scientific community in this multi-disciplinary area. Announcements for the Neural Interfaces Conference to be held in Spring 2008 will be forthcoming.

Neural interface technology has made enormous strides in recent years but stimulating electrodes remain incapable of reliably targeting specific cell types (e.g. excitatory or inhibitory neurons) within neural tissue. This obstacle has major scientific and clinical implications. For example, there is intense debate among physicians, neuroengineers and neuroscientists regarding the relevant cell types recruited during deep brain stimulation (DBS); moreover, many debilitating side effects of DBS likely result from lack of cell-type specificity. We describe here a novel optical neural interface technology that will allow neuroengineers to optically address specific cell types in vivo with millisecond temporal precision. Channelrhodopsin-2 (ChR2), an algal light-activated ion channel we developed for use in mammals, can give rise to safe, light-driven stimulation of CNS neurons on a timescale of milliseconds. Because ChR2 is genetically targetable, specific populations of neurons even sparsely embedded within intact circuitry can be stimulated with high temporal precision. Here we report the first in vivo behavioral demonstration of a functional optical neural interface (ONI) in intact animals, involving integrated fiberoptic and optogenetic technology. We developed a solid-state laser diode system that can be pulsed with millisecond precision, outputs 20 mW of power at 473 nm, and is coupled to a lightweight, flexible multimode optical fiber, ∼200 µm in diameter. To capitalize on the unique advantages of this system, we specifically targeted ChR2 to excitatory cells in vivo with the CaMKIIα promoter. Under these conditions, the intensity of light exiting the fiber (∼380 mW mm⁻²) was sufficient to drive excitatory neurons in vivo and control motor cortex function with behavioral output in intact rodents. No exogenous chemical cofactor was needed at any point, a crucial finding for in vivo work in large mammals. Achieving modulation of behavior with optical control of neuronal subtypes may give rise to fundamental network-level insights complementary to what electrode methodologies have taught us, and the emerging optogenetic toolkit may find application across a broad range of neuroscience, neuroengineering and clinical questions.

Prosthetic devices can be controlled using signals recorded in parts of the body where sensation and/or voluntary movement have been retained. Although neural prosthetic applications have used single-channel recordings, multiple-channel recordings could provide a significant increase in useable control signals. Multiple control signals can be acquired from recordings of a single implant by using a multi-contact electrode placed over a multi-fasciculated peripheral nerve. These recordings can be separated to recover the individual fascicular signals. Blind source separation (BSS) algorithms have been developed to extract independent source signals from recordings of their mixtures. The hypothesis that BSS algorithms can recover individual fascicular signals from nerve cuff recordings at physiological signal-to-noise ratio (SNR ∼ 3–10 dB) was investigated in this study using a finite-element model (FEM) of a beagle hypoglossal nerve with a flattening interface nerve electrode (FINE). Known statistical properties of fascicular signals were used to generate a set of four sources from which the neural signals recorded at the surface of the nerve with a multi-contact FINE were simulated. Independent component analysis (ICA) was then implemented for BSS of the simulated recordings. A novel post-ICA processing algorithm was developed to solve ICA's inherent permutation ambiguities. The similarity between the estimated and original fascicular signals was quantified by calculating their correlation coefficients. The mean values of the correlation coefficients calculated were higher than 0.95 (n = 50). The effects of the geometric layout of the FINE electrode and noise on the separation algorithm were also investigated. The results show that four distinct overlapping fascicular source signals can be simultaneously recovered from neural recordings obtained using a FINE with five or more contacts at SNR levels higher than 8 dB making them available for use as control signals.

Functional electrical stimulation (FES) holds great potential for restoring motor functions after brain and spinal cord injury. Currently, most FES systems are under simple finite state control, using external sensors which tend to be bulky, uncomfortable and prone to failure. Sensory nerve signals offer an interesting alternative, with the possibility of continuous feedback control. To test feasibility, we recorded from ensembles of sensory neurons with microelectrode arrays implanted in the dorsal root ganglion (DRG) of walking cats. Limb position and velocity variables were estimated accurately (average R² values >0.5) over a range of walking speeds (0.1–0.5 m s⁻¹) using a linear combination of firing rates from 10 or more neurons. We tested the feasibility of sensory control of intraspinal FES by recording from DRG neurons during hindlimb movements evoked by intraspinal microstimulation of the lumbar spinal cord in an anesthetized cat. Although electrical stimulation generated artifacts, this problem was overcome by detecting and eliminating events that occurred synchronously across the array of microelectrodes. The sensory responses to limb movement could then be measured and decoded to generate an accurate estimate of the limb state. Multichannel afferent recordings may thus provide FES systems with the feedback needed for adaptive control and perturbation compensation, though long-term stability remains a challenge.

Spinal cord injury (SCI) is a devastating neurological trauma that is prevalent predominantly in young individuals. Several interventions in the areas of neuroregeneration, pharmacology and rehabilitation engineering/neuroscience are currently under investigation for restoring function after SCI. In this paper, we focus on the use of neuroprosthetic devices for restoring standing and ambulation as well as improving general health and wellness after SCI. Four neuroprosthetic approaches are discussed along with their demonstrated advantages and their future needs for improved clinical applicability. We first introduce surface functional electrical stimulation (FES) devices for restoring ambulation and highlight the importance of these devices for facilitating exercise activities and systemic physiological activation. Implanted muscle-based FES devices for restoring standing and walking that are currently undergoing clinical trials are then presented. The use of implanted peripheral nerve intraneural arrays of multi-site microelectrodes for providing fine and graded control of force during sit-to-stand maneuvers is subsequently demonstrated. Finally, intraspinal microstimulation (ISMS) of the lumbosacral spinal cord for restoring standing and walking is introduced and its results to date are presented. We conclude with a general discussion of the common needs of the neuroprosthetic devices presented in this paper and the improvements that may be incorporated in the future to advance their clinical utility and user satisfaction.

Stem cells carry the promise to cure a broad range of diseases and injuries, from diabetes, heart and muscular diseases, to neurological diseases, disorders and injuries. Significant progresses have been made in stem cell research over the past decade; the derivation of embryonic stem cells (ESCs) from human tissues, the development of cloning technology by somatic cell nuclear transfer (SCNT) and the confirmation that neurogenesis occurs in the adult mammalian brain and that neural stem cells (NSCs) reside in the adult central nervous system (CNS), including that of humans. Despite these advances, there may be decades before stem cell research will translate into therapy. Stem cell research is also subject to ethical and political debates, controversies and legislation, which slow its progress. Cell engineering has proven successful in bringing genetic research to therapy. In this review, I will review, in two examples, how investigators are applying cell engineering to stem cell biology to circumvent stem cells' ethical and political constraints and bolster stem cell research and therapy.

The study of EEG recordings during the interval prior to an epileptic seizure onset—the preictal period—is likely to detect changes in the ongoing brain activity consistent with seizure anticipation. A novel index of spectral instability (ISpI) based on multiple abrupt changes of EEG spectral features is presented here. Based on the analysis of control records, robust M-estimates are used to calculate the threshold and avoid false warnings. The results obtained with a small data set (three patients, ten preictal records per patient) have shown that the ISpI index provided a warning flag that anticipated the seizure onset by 13.1 (SD = 4.0) min on average.

Neural models are increasingly being used as design components of physical systems. In order to most effectively utilize neuronal models in these novel contexts, we need to develop design rules for neuronal systems that relate how model design affects overall system performance. In this paper and a companion article, we investigate how the complexity of a neural model affects the performance of a two-cell oscillator built from the model. In this paper, we create a series of related neuron models with different mathematical complexity. Starting with a complex mechanistic model of a bursting neuron, we use a variety of techniques to create a series of simplified neuron models. These three reduced models produce bursting activity that is qualitatively very similar to the original model. In the following companion article, we investigate the functional performance of oscillators built from these models.

Neural models are increasingly being used as design components of physical systems. In order to best use models in these novel contexts, we must develop design rules that describe how decisions in model construction relate to the functional performance of the resulting system. In the accompanying paper, we described a series of related neuron models of varying complexity. Here, we use these models to build several half-center oscillators, and investigate how model complexity influences the robustness and flexibility of these oscillators. Our results indicate that model complexity has a significant effect on the robustness and flexibility of systems that incorporate neural models.

Holes in the skull and the scalp are associated with intracranial monitoring procedures. The purpose of the present study is to evaluate the effects of holes on extracranial electroencephalogram (EEG) and intracranial electrocorticogram (ECoG) recordings. The finite difference method (FDM) was used to model the head volume conductor with a hole of varying size. A current dipole was used to simulate the brain electrical activity with varying locations within the brain. The effects of the holes were assessed by comparing the forward potential distributions with and without a hole. The present computer simulation results indicate that the effect of a hole on the scalp EEG and ECoG recordings strongly depends on the dipole location and orientation. For a superficial radial dipole located under a hole of radius ranging from 5 mm to 40 mm, the relative error (RE) varies from 0.99% to 93.07% for the EEG and from 0.025% to 16.72% for the ECoG. The correlation coefficient (CC) varies from 99.99% to 21.1% and from 100% to 99.75% for the EEG and EcoG, respectively. For radial dipoles, the strongest effect on the EEG and ECoG occurs when the dipole is located below the center of the hole, while for tangential dipoles, the strongest effect occurs when the dipole is located below the border of the hole. The effect of a hole on the EEG is much larger than upon the ECoG.

The stimulation of the vagus nerve has been used as an anti-epileptic treatment for over a decade, and its use for depression and chronic heart failure is currently under investigation. Co-activation of the intrinsic laryngeal muscles may limit the clinical use of vagal stimulation, especially in the case of prolonged activation. To prevent this, the use of a selective stimulation paradigm has been tested in seven acute pig experiments. Quasi-trapezoidal pulses successfully blocked the population of the largest and fastest vagal myelinated fibers being responsible for the co-activation. The first response in the vagus compound action potential was reduced by 75 ± 22% (mean ± SD) and the co-activated muscle action potential by 67 ± 25%. The vagal bradycardic effects remained unchanged during the selective block, confirming the leading role of thin nerve fibers for the vagal control of the heart. Quasi-trapezoidal pulses may be an alternative to rectangular pulses in clinical vagal stimulation when the co-activation of laryngeal muscles must be avoided.

Recent advances in functional electrical stimulation (FES) show significant promise for restoring voluntary movement in patients with paralysis or other severe motor impairments. Current approaches for implantable FES systems involve multisite stimulation, posing research issues related to their physical size, power and signal delivery, surgical and safety challenges. To explore a different means for delivering the stimulus to a distant muscle nerve site, we have elicited in vitro FES response using a high efficiency microcrystal photovoltaic device as a neurostimulator, integrated with a biocompatible glass optical fiber which forms a lossless, interference-free lightwave conduit for signal and energy transport. As a proof of concept demonstration, a sciatic nerve of a frog is stimulated by the microcrystal device connected to a multimode optical fiber (core diameter of 62.5 µm), which converts optical activation pulses (∼100 µs) from an infrared semiconductor laser source (at 852 nm wavelength) into an FES signal.

A brain–computer interface (BCI) is a device that allows a user to communicate with external devices through thought processes alone. A novel signal acquisition tool for BCIs is near-infrared spectroscopy (NIRS), an optical technique to measure localized cortical brain activity. The benefits of using this non-invasive modality are safety, portability and accessibility. A number of commercial multi-channel NIRS system are available; however we have developed a straightforward custom-built system to investigate the functionality of a fNIRS-BCI system. This work describes the construction of the device, the principles of operation and the implementation of a fNIRS-BCI application, 'Mindswitch' that harnesses motor imagery for control. Analysis is performed online and feedback of performance is presented to the user. Mindswitch presents a basic 'on/off' switching option to the user, where selection of either state takes 1 min. Initial results show that fNIRS can support simple BCI functionality and shows much potential. Although performance may be currently inferior to many EEG systems, there is much scope for development particularly with more sophisticated signal processing and classification techniques. We hope that by presenting fNIRS as an accessible and affordable option, a new avenue of exploration will open within the BCI research community and stimulate further research in fNIRS-BCIs.

Historically the rectangular pulse waveform has been the choice for neural stimulation. The strength–duration curve is thus defined for rectangular pulses. Not much attention has been paid to alternative waveforms to determine if the pulse shape has an effect on the strength–duration relation. Similarly the charge injection capacity of neural electrodes has also been measured with rectangular pulses. In this study we questioned if non-rectangular waveforms can generate a stronger stimulation effect, when applied through practical electrodes, by minimizing the neural activation threshold and maximizing the charge injection capacity of the electrode. First, the activation threshold parameters were studied with seven different pulse shapes using computer simulations of a local membrane model. These waveforms were rectangular, linear increase and decrease, exponential increase and decrease, Gaussian, and sinusoidal. The chronaxie time was found to be longer with all the non-rectangular pulses and some provided more energy efficient stimulation than the rectangular waveform. Second, the charge injection capacity of titanium nitride microelectrodes was measured experimentally for the same waveforms. Linearly decreasing ramp provided the best charge injection for all pulse widths tested from 0.02 to 0.5 ms. Finally, the most efficient waveform that maximized the charge injection capacity of the electrode while providing the lowest threshold charge for neural activation was searched. Linear and exponential decrease, and Gaussian waveforms were found to be the most efficient pulse shapes.

Maintaining the postural configuration of a limb such as an arm or leg is a fundamental neural control task that involves the coordination of multiple linked body segments. Biological systems are known to use a complex network of inter- and intra-joint feedback mechanisms arising from muscles, spinal reflexes and higher neuronal structures to stabilize the limbs. While previous work has shown that a small amount of asymmetric heterogenic feedback contributes to the behavior of these systems, a satisfactory functional explanation for this non-conservative feedback structure has not been put forth. We hypothesized that an asymmetric multi-joint control strategy would confer both an energetic and stability advantage in maintaining endpoint position of a kinematically redundant system. We tested this hypothesis by using optimal control models incorporating symmetric versus asymmetric feedback with the goal of maintaining the endpoint location of a kinematically redundant, planar limb. Asymmetric feedback improved endpoint control performance of the limb by 16%, reduced energetic cost by 21% and increased interjoint coordination by 40% compared to the symmetric feedback system. The overall effect of the asymmetry was that proximal joint motion resulted in greater torque generation at distal joints than vice versa. The asymmetric organization is consistent with heterogenic stretch reflex gains measured experimentally. We conclude that asymmetric feedback has a functionally relevant role in coordinating redundant degrees of freedom to maintain the position of the hand or foot.

A brain–computer interface (BCI) provides a communication channel that translates human intention reflected by a brain signal such as electroencephalogram (EEG) into a control signal for an output device. In this work, the main concern is to reduce the training effort for BCI, which is often tedious and time consuming. Here we introduce a transductive support vector machines (TSVM) algorithm for the classification of EEG signals associated with mental tasks. TSVM possess the property of using both labeled and unlabeled data for reducing the calibration time in BCI and achieving good performance in classification accuracy. The advantages of the proposed method over the traditional supervised support vector machines (SVM) method are confirmed by about 2%–9% higher classification accuracies on a set of EEG recordings of three subjects from three-tasks-based mental imagery experiments.

Progenitor and stem cell transplantation represent therapeutic strategies for retinal disorders that are accompanied by photoreceptor degeneration. The transplanted cells may either replace degenerating photoreceptors or secrete beneficial factors that halt the processes of photoreceptor degeneration. The present study analyzes whether rat retinal progenitor cells differentiated into photoreceptor phenotypic cells in neurospheres have a potential to interact with rat retinal explants. Immunocytochemistry for rhodopsin and synaptophysin indicated photoreceptor cell-like differentiation in neurospheres that were stimulated by basic fibroblast growth factor and epidermal growth factor. Differentiation into neural phenotypes including photoreceptor cells was effectively blocked by an addition of leukemia inhibitory factor. Grafting of neurospheres onto retinal explants demonstrated a consistent penetration of glial cell processes into the explanted tissue. On the other hand, the incorporation of donor cells into explants was very low. A general finding was that neurospheres grafting was associated with local decrease in Müller cell activation in the explants. Further characterization of these effect(s) could provide further insight into progenitor cell-based therapies of retinal degenerative disorders.

Signals from the brain could provide a non-muscular communication and control system, a brain–computer interface (BCI), for people who are severely paralyzed. A common BCI research strategy begins by decoding kinematic parameters from brain signals recorded during actual arm movement. It has been assumed that these parameters can be derived accurately only from signals recorded by intracortical microelectrodes, but the long-term stability of such electrodes is uncertain. The present study disproves this widespread assumption by showing in humans that kinematic parameters can also be decoded from signals recorded by subdural electrodes on the cortical surface (ECoG) with an accuracy comparable to that achieved in monkey studies using intracortical microelectrodes. A new ECoG feature labeled the local motor potential (LMP) provided the most information about movement. Furthermore, features displayed cosine tuning that has previously been described only for signals recorded within the brain. These results suggest that ECoG could be a more stable and less invasive alternative to intracortical electrodes for BCI systems, and could also prove useful in studies of motor function.

Glutamate spillover is thought to play a significant role in increasing neural transmission at the mossy fiber/granule cell cerebellar glomerulus. Glutamate spillover has been shown to activate AMPA receptors at the glomerulus, and here we complete the characterization of spillover at the glomerulus by investigating the role of glutamate spillover in N-methyl-D-aspartate receptor (NMDA-R) activation. We present a quantitative model of glutamate spillover combining recent models of glutamate diffusion and NMDA-R binding to determine the open probabilities of NMDA-Rs over time at a neighbor synapse. Simulation results from a baseline set of physiologically realistic parameters show that glutamate spillover onto a single neighbor synapse, created by glutamate that diffuses from a point source into a restricted fractional 2D–3D space and the glutamate concentration created by neighboring glutamate release sites, is sufficient to elicit an NMDA-R peak open probability of 0.23, approximately 79% of that obtained by a direct release (peak open probability of 0.29). Thus, it would appear that glutamate spillover at the glomerulus at NMDA receptors is even more substantial than that seen at AMPA receptors.

Time-varying magnetic fields can induce electric fields in the neuronal tissue, a phenomenon that has been recently explored in clinical applications such as peripheral nerve stimulation and transcranial magnetic stimulation. Although the transmembrane potential induced during direct electric stimulation has already been the subject of a number of theoretical studies, an analytical solution for the magnetically induced transmembrane potential change is still unavailable. In addition, although several studies have analyzed the impact of stimulation parameters, including stimulation intensity and frequency, as well as coil design and position, on the amount of tissue polarization, the effects of tissue non-homogeneity on cell polarization have not been fully elucidated. In this study, we have derived an analytical expression for the transmembrane potential induced by a low-frequency magnetic field in a spherical neuronal structure. This model is representative of a spherical cell body or any neuronal structure of a similar shape. The model cell is located in an extracellular medium and possesses a low-conductive membrane and an internal cytoplasm. These three regions represent the basic tissue non-homogeneity of a neuron at a microscopic level. The sensitivity of the induced transmembrane potential to the coil position and to the geometrical and electrical parameters of the model structure was studied in a broad physiologically relevant range. Our results demonstrate that the structure is regionally polarized, with the pattern of polarization depending on the relative positioning between the model cell and the stimulation coil. In addition, both the geometrical and electrical parameters of the structure affect the amount of polarization. These results may be generalized to other neuronal tissues that possess similar non-homogenous properties, but different shapes, such as an axon. Our results support the idea that aside from coil design and position, tissue non-homogeneity could play an important role in determining the effects of magnetic stimulation.

Electrically interfaced cortical networks cultured in vitro can be used as a model for studying the network mechanisms of learning and memory. Lasting changes in functional connectivity have been difficult to detect with extracellular multi-electrode arrays using standard firing rate statistics. We used both simulated and living networks to compare the ability of various statistics to quantify functional plasticity at the network level. Using a simulated integrate-and-fire neural network, we compared five established statistical methods to one of our own design, called center of activity trajectory (CAT). CAT, which depicts dynamics of the location-weighted average of spatiotemporal patterns of action potentials across the physical space of the neuronal circuitry, was the most sensitive statistic for detecting tetanus-induced plasticity in both simulated and living networks. By reducing the dimensionality of multi-unit data while still including spatial information, CAT allows efficient real-time computation of spatiotemporal activity patterns. Thus, CAT will be useful for studies in vivo or in vitro in which the locations of recording sites on multi-electrode probes are important.

A fully implantable neural data acquisition system is a key component of a clinically viable cortical brain–machine interface. We present the design and implementation of a single-chip device that serves the processing needs of such a system. Our device processes 96 channels of multi-unit neural data and performs all digital processing necessary for bidirectional wireless communication. The implementation utilizes a single programmable logic device that is responsible for performing data reduction on the 96 channels of neural data, providing a bidirectional telemetry interface to a transceiver and performing command interpretation and system supervision. The device takes as input neural data sampled at 31.25 kHz and outputs a line-encoded serial bitstream containing the information to be transmitted by the transceiver. Data can be output in one of the following four modes: (1) streaming uncompressed data from a single channel, (2) extracted spike waveforms from any subset of the 96 channels, (3) 1 ms bincounts for each channel or (4) streaming data along with extracted spikes from a single channel. The device can output up to 2000 extracted spikes per second with latencies suitable for a brain–machine interface application. This device provides all of the digital processing components required by a fully implantable system.

Spike sorting is a technologically expensive component of the signal processing chain required to interpret population spike activity acquired in a neuromotor prosthesis. No systematic analysis of the value of spike sorting has been carried out, and little is known about the effects of spike sorting error on the ability of a brain–machine interface (BMI) to decode intended motor commands. We developed a theoretical framework to examine the effects of spike processing on the information available to a BMI decoder. We computed the mutual information in neural activity in a simplified model of directional cosine tuning to compare the effects of pooling activity from up to four neurons to the effects of sorting with varying amounts of spike error. The results showed that information in a small population of cosine-tuned neurons is maximized when the responses are sorted and there is diverse tuning of units, but information was affected little when pooling units with similar preferred directions. Spike error had adverse effects on information, such that non-sorted population activity had 79–92% of the information in its sorted counterpart for reasonable amounts of detection and sorting error and for units with moderate differences in preferred direction. This quantification of information loss associated with pooling units and with spike detection and sorting error will help to guide the engineering decisions in designing a BMI spike processing system.

Neural prostheses aim to improve the quality of life of severely disabled patients by translating neural activity into control signals for guiding prosthetic devices or computer cursors. We recently demonstrated that plan activity from premotor cortex, which specifies the endpoint of the upcoming arm movement, can be used to swiftly and accurately guide computer cursors to the desired target locations. However, these systems currently require additional, non-neural information to specify when plan activity is present. We report here the design and performance of state estimator algorithms for automatically detecting the presence of plan activity using neural activity alone. Prosthesis performance was nearly as good when state estimation was used as when perfect plan timing information was provided separately (∼5 percentage points lower, when using 200 ms of plan activity). These results strongly suggest that a completely neurally-driven high-performance brain–computer interface is possible.