Author: Anderson, Nicholas R
Date published: September 1, 2010
A brain computer interface (BCI) is a system that bypasses the nervous system to allow control of external devices using signals generated by the brain. While BCIs can take on many forms, the universal goal of BCIs is to allow a patient to regain control of, or provide a substitute for, a function they have lost, (i.e., allow them to communicate with the outside world). Depending on the level of complexity, these functions can range from speech and simple typing tasks to control of a prosthetic hand.
A typical BCI (Figure 1) records signals from the brain, interprets them, and then affects some change in a computer or device. The primary motivation for the creation of these devices is to help those who suffer from neurological injury as a result of stroke and other neurological disorders [e.g., amyotropic lateral sclerosis (ALS)]. Each of these diseases and injuries has a different onset mechanism, but the resulting paralysis is universal (AHA 2010, NSCIA 2007). Due to the nature of these different mechanisms, curing these conditions through biological means would require a myriad of treatments. BCI provides a possible solution for all paralysis patients regardless of their specific disease.
Currently, most applications of BCIs are in the research phase, with only a few modalities making their way into clinical applications. EEG in particular, having benefited from many years of signal process work and computer programming, has made marked differences in the treatment of patients who are fully cognizant of their surroundings, but unable to move or communicate (i.e., "locked in"), allowing them to control a computer in multiple ways for communication (Wolpaw et al. 1991, Wolpaw and McFarland 2004, McFarland et al. 2005, Birbaumer 2006, Schwartz et al. 2006). Along with scalp EEG, subdural EEG (which places electrodes directly on the surface of the brain) (Figure 2) will allow for the creation of more robust and intuitive BCIs. With better signal quality than scalp EEG and the ability to be learned in a matter of minutes, subdural EEG has been successfully employed in both research and clinical settings for use with computer programs as well as having limited applications in BCI (Schwartz et al. 2004, Rektor et al. 2006, Schalk et al. 2007).
Typically BCIs are made up of three distinct components. The first is a recording device such as an EEG that enables the reading of cortical signals. Once recorded, the cortical signal is analyzed by a computer to recognize specific features of the signal, such as a spike in amplitude, that denote the patient's desired movement (Leuthardt et al. 2004). Since each anticipated movement and subsequent cortical signal corresponds to a distinct signal feature, the computer is able to use a pre-programmed algorithm to translate this feature into a command every time the feature is recognized (Birbaumer 2006). This command is then relayed to an external device such as a prosthetic or computer that performs the desired function (e.g., speech, movement). Following this model, BCIs allow humans suffering from a stroke or other severe physiological problems to bypass their central nervous system, thereby executing movements or sounds using only a computer and their brain.
The work to create a useful BCI as a human prosthetic began only 30 years ago with the help of EEG (Vidal 1973). While it seemed that creating an EEG based BCI would be a relatively simple endeavor, many practical and engineering problems were encountered (e.g., excessive noise and long user training times), belaboring the development of a pragmatic BCI for those with immediate need (Wolpaw et al. 2003, Schalk 2005). In response to these hurdles, BCIs have progressed seemingly independently in non-human primates and humans because of the vast difference in recording paradigms.
The most common recording methodology for cortical activation studies in humans is functional magnetic resonance imaging (fMRI), which uses MRI to map brain function as signals occur. These experiments, when performed on humans, tend to show less specific results than related non-human primate literature, but give a better picture of overall brain activity (Matelli et al. 1993, Moran and Schwartz 1999). While research using fMRI continues, the high cost of equipment and relatively low temporal resolution make fMRI an unlikely candidate for clinical applications of BCI (Wolpaw et al. 2002).
Historically, control signals - stable signals capable of being used for BCIs - have been extracted from the human cortex using EEG, which has been the basis for developing human BCIs (Wolpaw and Birbaumer 2006). Placed on the scalp, EEG is a safe and inexpensive technique that allows for a substantial amount of information to be recorded at one time. With a large number of labs running this paradigm, the years of signal process work and computer programming that have been invested in EEG have made it the most successful modality for human BCIs at the present time (Wolpaw et al. 1 99 1 , Wolpaw and McFarland 2004, McFarland et al. 2005, Birbaumer 2006). As such, most currently available clinical BCIs are based on EEG. However, despite EEG's success as a BCI platform, it has some limitations such as a slow learning curve (greater than one hour) and a high amount of recorded noise, that keep it from realizing the level of success initially hoped for.
Magnetoencephalogram (MEG) is another non-invasive technique used in BCI research. Frontal cortex has been the focus of many different studies in MEG, including those for motor related tasks (Sasaki et al. 1996, Kawaguchi et al. 2005, Nakata et al. 2005). These studies have focused on simple go/no-go related signals and not on possible directional tuning or target encoding (i.e., signals that encode nonspecific commands such as "move" or "don't move" with no details about direction or when to stop the movement). Though MEG has reduced training times and is functionally similar to EEG, MEG has been less successful due to the high cost of equipment and high degree of noise present in the recordings (Mellinger et al. 2005). MEG has, therefore, been limited to applications in BCI research, making EEG the non-invasive focus of human BCI research.
Developed largely in parallel to the EEG interfaces, microelectrodes, which are, surgically placed directly into the brain tissue, have proven to be a successful recording modality for BCIs as well (Fetz and Finocchio 1971, Schwartz et al. 1988, Georgopoulos 1995, Nicolelis et al. 2003). Studies performed using microelectrodes have yielded many significant results in the study of motor regions, beginning with the demonstration that a monkey's arm movements are correlated with action potentials (Polit and Bizzi 1979). Building upon this result, it was shown that hand direction is represented in the motor cortex (Georgopoulos et al. 1986) and that arm position is encoded by a subpopulation of motor cortical neurons (Evarts et al. 1983). Based on this research BCIs have been developed and successfully used by nonhuman primates in research settings. However, despite the more detailed information obtained by microelectrodes (Fetz and Finocchio 1975), control using microelectrodes has primarily been limited to non-human primates and other animals because of their questionable long term stability and the damage to brain tissue that occurs upon insertion. This instability of single units (their tendency to move within the brain tissue over time) (Kipke et al. 2003, Vetter et al. 2004) has caused interest to shift to field potentials (similar to subdural EEG) of small groups of neurons that may provide the signal quality present in single units, while also offering a level of safety and stability suitable for human use (Heldman et al. 2006, Hochberg et al. 2006).
Field potentials, recorded by subdural EEG, currently show great promise for long term encoding of brain activity for use in brain computer interfaces (Andersen et al. 2004a, Andersen et al. 2004b, Schwartz et al. 2006). Along with being more robust and less technically demanding than recording action potentials directly, as is done with single units (Pesaran et al. 2002, Heldman et al. 2006), field potentials have been shown to correlate to a variety of motor and cognitive activities that show great potential for use with BCIs (Crone et al. 2001, Schalk et al. 2007). More specifically, subdural EEG based BCI has shown the ability to allow the user to learn to control a computer with just a few minutes of training.
Although subdural EEG has proven to be more robust in clinical settings, it has not yet been shown to decode more degrees of freedom, which corresponds directly with level of complexity of the BCI, than its EEG counterpart (Leuthardt et al. 2004, Schalk 2005). Degrees of freedom are short hand in the BCI community for the number of independent control signals a patient has, controlling a computer requires three degrees of freedom, moving in a 3-dimensional environment requires three degrees of freedom, controlling a hand requires five degrees of freedom (one for each finger). Furthermore, the noninvasive nature of EEG-based BCIs allow BCIs to reach a broader spectrum of patients than more invasive alternatives. Due to its many benefits, state of the art EEG BCI systems have recently been moved from their laboratory roots into clinical environments. The improved functionality from these clinically developed BCIs allow some of them to be administrated on a daily basis by the patient's nurse (Birbaumer 2006).
The first step in creating a usable BCI system is acquisition of a viable cortical signal. Typically, this signal is acquired via one of four commonly used methods, including the electroencephalogram (EEG), subdural EEG (macro local field potentials), microelectrodes (single units), and hybrid electrodes (micro subdural EEG). Other recording methods less commonly employed for BCI use include magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI). These methods vary greatly in level of invasiveness as well as spectral/temporal resolution of the signal.
EEG makes use of strategically placed scalp electrodes to record signals from the brain. Due to the extracranial location of the electrodes, the impulse of a single neuron cannot be read. Instead, the acquired signal is an average of large groups of neurons firing in close proximity to a given electrode (Fetz and Finocchio 1975). Though this neuronal averaging tends to present lower signal fidelity than other more invasive forms of signal acquisition, this non-invasive technique is successfully and routinely employed for use with BCIs (Birbaumer 2006).
Subdural EEG [Electrocorticography (ECoG), macro local field potentials]
Subdural EEG requires surgical implantation of the electrodes directly onto the surface of the brain. Though substantially more invasive than EEG, subdural EEG recordings allow for more focused acquisition of neuronal signals and substantially less spatial averaging (Leuthardt et al. 2004, Wisneski et al. 2008, Anderson et al. 2009). This technique also allows for the recording of field potentials, which have been shown to carry information about both synaptic and action potentials (Heldman et al. 2006). Furthermore, subdural EEG has a much better signal to noise ratio at higher frequencies (Leuthardt et al. 2004). Due to its invasive nature, subdural EEG is mainly employed in research settings on epilepsy patients who, for medical reasons, require subdural electrodes to monitor seizure activity.
Microelectrodes (Single Units, high impedance, local field potentials)
Implanted directly into cortical tissue, microelectrodes provide highly detailed information per electrode because they record many independent neurons rather than an average of these same neurons as a large group (Fetz and Finocchio 1975, Georgopoulos et al. 1982, Georgopoulos et al. 1986, Schwartz et al. 1988, Moran and Schwartz 1999). Since action potentials of a single neuron provide the highest signal fidelity, BCIs utilizing microelectrodes will theoretically provide the most useful control (Nunez and Srinivasan 2006). However, while widely used in animal models, microelectrodes are limited in their human application because they both damage the cortex upon insertion and have uncertain long-term stability (Kipke et al. 2003, Vetter et al. 2004, Heldman et al. 2006).
Hybrid Electrodes (micro subdural EEG)
Hybrid electrodes make use of a scaled down version of subdural EEG electrodes, and currently are primarily used for clinical research. Identical to subdural EEG, hybrid electrodes must be surgically implanted on the surface of the cortex and are only used for research with patients who require the electrodes for medical purposes. The benefit of hybrid electrodes is that they record cortical signals over a smaller area, requiring less spatial averaging and therefore providing higher signal fidelity while still maintaining a less invasive approach than single units.
Magnetoencephalography (MEG) is closely related to EEG but it detects the magnetic component (rather than the electrical component) of the signal on the surface of the skull. Unlike EEG, which records signals primarily from the gyri of the cortex, MEG records primarily from the sulci (Cohen 1968, Zimmerman et al. 1970). It is also a technique that allows for much better source localization than EEG because it does not depend on head geometry, as the magnetic signals are not randomly scattered like the EEG signals. MEG as a BCI modality has the advantage of better signal to noise ratio and better source localization than EEG; however, due to its high cost it is less widely used than many cheaper, but invasive methods (Leuthardt et al. 2004).
Functional Magnetic Resonance Imaging (fMRI)
Though functional magnetic resonance imaging (fMRI) is the most common recording method for cortical activation studies in humans, the high cost of machinery has been a limiting factor in BCI applications. In general, fMRI yields better spatial resolution than any other non-invasive modality, and can also record activity from the whole brain simultaneously, which is not available with any other modality (Rao et al. 1 993, Cohen and Bookheimer 1994). However, fMRI studies have not been widely used in decoding the activations in motor cortex because of the difficulty in recording movements in fMRI machine space (Gassert et al. 2005). Therefore, despite its success in other areas, the expense of the equipment and the relatively low temporal resolution make fMRI an unlikely candidate for practical applications of BCI, compared to less expensive alternatives that can provide comparatively good results (Wolpaw et al. 2002).
The complexity and usefulness of clinical applications of BCI varies directly with the quantity and quality of information gathered from the cortical signals. So while many of the clinical applications are obvious they often require very sophisticated BCIs to operate them. Recording methods that entail a high degree of spatial averaging such as EEG and fMRI can only control BCIs that only allow for simple movements in one plane, while highly specific modalities such as microelectrodes can allow for movements along multiple axes. In general, the more robust and specific the signal, the higher the degree of control the BCI can offer the patient.
As shown in Table 1 , clinical applications of BCIs range from simple tasks operable in one plane, such as allowing the patients to raise and lower their own bed, to fully restorative tasks that allow the patients to regain the complete range of motion in part, or all, of their body.
Along with purely functional applications, the advances being made in BCI research have the potential to aid both researchers and clinicians as BCIs expand into more clinical applications in the coming years. One key component of BCIs that will be of great use to clinicians is the advanced signal processing BCIs use to decode brain signals. The promise for this technology lies in its potential to decode disease states in the human brain. Currently, neurological diseases are hard to diagnose and classify unless specific symptoms are present. However, BCI research has shown that the cortical changes decoded for BCI are much smaller than those in many disease states (Leuthardt et al. 2006). This means that as small changes occur within the brain leading to neurological impairment, these recording mechanisms will be able to detect them far earlier than ever before. Thus, with minor changes to the BCI algorithms these processing mechanisms will make it possible for clinicians to detect disease states using concrete means, rather than waiting for possible symptoms to present themselves.
In addition, BCI research is creating useful tools for mapping the cortex. Cortex mapping is typically used in BCI research to track where key activations take place, as well as to position the electrodes on the brain. These mapping tools can create maps of the brain much faster than standard clinical practices (Leuthardt et al. 2006, Miller et al. 2007, Anderson et al. 2009). More specifically, many of these mapping methods are agnostic signal processing techniques that can be used in many different ways, including mapping for clinical purposes. These techniques are beginning to be explored for use in epilepsy and brain tumor mapping, and, if successful, would greatly aid doctors in diagnosing and treating patients (Schalk et al. 2008a, Schalk et al. 2008b, Brunner et al. 2009, Darvas et al. 2010).
Other useful tools under investigation by BCI researchers include automation of stereotactic locations on the brain and tools for simultaneously stimulating subjects with visual cues while recording EEG signals (Schalk et al. 2004, Miller et al. 2007, Schalk 2009, Hermes et al. 2010). Both of these techniques have potential in detecting and diagnosing diseases and will also greatly benefit researchers in other fields of interest. These general methods help to allow cortical mapping in all clinical indications where it is needed and help to focus clinical investigations. The tools that are being created by BCI researchers have clinical utility outside of their intended BCI applications and will impact clinical care in new ways in the near future.
American Heart Association (AHA). Heart disease and stroke statistics. 2010. On the Internet at: http://www.americanheart.org/downloadable/heart/ 1 265665 1 52970DS-324 1 %20HeartStrokeUpdate_2010.pdf Accessed March 2010.
Andersen RA, Burdick JW, Musallam S, Scherberger H, Pesaran B, Meeker D, Cornell BD, Fineman I, Nenadic Z, Branchaud E, Chan JG, Greger B, Tai YC, Mojarradi MM. Recording advances for neural prosthetics. Conf Proc IEEE Eng Med Biol Soc 2004a; 7:5352-55.
Andersen RA, Musallam S, Pesaran B. Selecting the signals for a brain-machine interface. Curr Opin Neurobiol 2004b; 14:720-26.
Anderson NR, Wisneski K, Eisenman L, Moran DW, Leuthardt EC, Krusienski DJ. An offline evaluation of the autoregressive spectrum of electrocorticography. IEEE Trans Biomed End 2009; 56:913-16.
Birbaumer N. Brain-computer-interface research: coming of age. Clin Neurophysiol 2006; 117:479-83.
Brunner P, Ritaccio AL, Lynch TM, Emrich JF, Wilson JA, Williams JC, Aarnoutse EJ, Ramsey NF, Leuthardt EC, Bischof H, Schalk G. A practical procedure for real-time functional mapping of eloquent cortex using electrocorticographic signals in humans. Epilepsy Behav 2009, 15: 278-86.
Cohen D. Magnetoencephalography: evidence of magnetic fields produced by alpha-rhythm currents. Science 1968; 161:784-86.
Cohen MS, Bookheimer SY. Localization of brain function using magnetic resonance imaging. Trends Neurosci 1994; 17:268-77.
Crone NE, Hao L, Hart J Jr, Boatman D, Lesser RP, Irizarry R, Gordon B. Electrocorticographic gamma activity during word production in spoken and sign language. Neurology 2001; 57:2045-53.
Darvas F, Scherer R, Ojemann JG, Rao RP, Miller KJ, Sorensen LB. High gamma mapping using EEG. Neuroimage 2010; 49:930-38.
Evarts EV, Fromm C, Kröller J, Jennings VA. Motor cortex control of finely graded forces. J Neurophysiol 1 983; 49: 1 1 99-1 2 1 5.
Fetz EE, Finocchio DV. Operant conditioning of specific patterns of neural and muscular activity. Science 1971; 174:431-35.
Fetz EE, Finocchio DV. Correlations between activity of motor cortex cells and arm muscles during operantly conditioned response patterns. Exp Brain Res 1975; 23:217-40.
Gassert R, Dovat L, Ganesh G, Bürdet E, Imamizu H, Milner T, Bleuler H. Multi-joint arm movements to investigate motor control with fMRI. Conf Proc IEEE Eng Med Biol Soc 2005; 5:4488-91.
Georgopoulos AP. Current issues in directional motor control. Trends Neurosci 1995; 18:506-10.
Georgopoulos AP, Kalaska JF, Caminiti R, Massey JT. On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex, J Neurosci 1982; 2:1527-37.
Georgopoulos AP, Schwartz AB, Kettner RE. Neuronal population coding of movement direction. Science 1986;233:1416-19.
Heldman D. Epidural electrocorticography and intra-cortical local field potentials in motor cortex during volitional arm movements and their application to neural prosthetic control. Biomedical Engineering 2007.
Heldman DA, Wang W, Chan SS, Moran DW. Local field potential spectral tuning in motor cortex during reaching. IEEE Trans Neural Syst Rehabil Eng 2006; 14: 1 80-83.
Hermes D, Miller KJ, Noordmans HJ, Vansteensel MJ, Ramsey NF. Automated electrocorticographic electrode localization on individually rendered brain surfaces. J Neurosci Methods 2010; 185:293-98.
Hochberg LR, Serruya MD, Friehs GM, Mukand JA, Saleh M, Caplan AH, Branner A, Chen D, Penn RD, Donoghue JP. Neuronal ensemble control of prosthetic devices by a human with tetraplegia. Nature 2006; 442:164-71.
Kawaguchi S, Ukai S, Shinosaki K, Ishii R, Yamamoto M, Ogawa A, Mizuno-Matsumoto Y, Fujita A, Yoshimine T, Takeda M. Information processing flow and neural activations in the dorsolateral prefrontal cortex in the Stroop task in schizophrenia patients. A spatially filtered MEG analysis with high temporal and spatial resolution. Neuropsycholobiology 2005; 51:191-203.
Kipke DR, Vetter RJ, Williams JC, Hetke JF. Silicon-substrate intracortical microelectrode arrays for long-term recording of neuronal spike activity in cerebral cortex. IEEE Trans Neural Syst Rehabil Eng 2003; 1 1 (2): 1 51-55.
Leuthardt EC, Schalk G, Wolpaw JR, Ojemann JG, Moran DW. A brain-computer interface using electrocorticographic signals in humans. J Neural Eng 2004 Jun; 1:63-71.
Leuthardt EC, Schalk G, Moran D, Ojemann JG. The emerging world of motor neuroprosthetics: a neurosurgical perspective. Neurosurgery 2006; 59: 1-14.
Leuthardt EC, Miller KJ, Anderson N, Schalk G, Dowling J, Moran D, Miller J, Ojemann JG. Electrocorticographic frequency alteration mapping (EFAM): a novel clinical technique for mapping motor cortex. Neurosurgery 2010. In press.
Matelli M, Rizzolatti G, Bettinardi V, Guardi MC, Perani D, Rizzo G, Fazio F. Activation of precentrai and mesial motor areas during the execution of elementary proximal and distal arm movements: a PET study. Neuroreport 1 993; 4: 1 295-98.
McFarland DJ, Sarnacki WA, Wolpaw JR. Reach and select with a non-invasive brain-computer interface in humans: emulating full mouse control. Presented at the Society for Neuroscience 2005; Washington, DC.
Mellinger J, Braun C, Preissl H, Schalk G, Birbaumer N, Kubier A. A brain-computer interface based on magneto-encephalography. Presented at the Society for Neuroscience 2005; Washington, DC.
Miller KJ, Leuthardt EC, Schalk G, Rao RP, Anderson NR, Moran DW, Miller JW, Ojemann JG. Spectral changes in cortical surface potentials during motor movement J Neurosci 2007; 27:2424-32.
Miller KJ, Makeig SH, Hebb AO, Rao RP, denNijs M, Ojemann JG. Cortical electrode localization from X-rays and simple mapping for electrocorticographic research: the 'location on cortex' (LOC) package for MATLAB. J Neurosci Methods 2007; 162:303-08.
Moran DW, Schwartz AB. Motor cortical representation of speed and direction during reaching. J Neurophysiol 1999; 82:2676-92.
Nakata H, Inui K, Wasaka T, Akatsuka K, Kakigi R. Somato-motor inhibitory processing in humans: a study with MEG and ERP. Eur J Neurosci 2005; 22: 1 784-92.
National Spinal Cord Injury Association (NSCIA). Spinal cord injury statistics. 2007. On the Internet at: www.spinalcord.org Accessed March 2010.
Nicolelis MA, Dimitrov D, Carmena JM, Crist R, Lehew G, Kralik JD, Wise SP. Chronic, multisite, multielectrode recordings in macaque monkeys. Proc Natl Acad Sci U S A 2003; 100: 1 1041-46.
Pesaran B, Pezaris JS, Sahani M, Mitra PP, Andersen RA. Temporal structure in neuronal activity during working memory in macaque parietal cortex. Nat Neurosci 2002; 5:805-1 1.
Polit A, Bizzi E. Characteristics of motor programs underlying arm movements in monkeys. J Neurophysiol 1979;42:183-94.
Rao SM, Binder JR, Bandettini PA, Hammeke TA, Yetkin FZ, Jesmanowicz A, Lisk LM, Morris GL, Mueller WM, Estkowski LD. Functional magnetic resonance imaging of complex human movements. Neurology 1993; 43:231 1-18.
Rektor I, Sochurková D, Bocková M. Intracerebral ERD/ERS in voluntary movement and in cognitive visuomotor task. Prog Brain Res 2006; 159:31 1-30.
Sasaki K, Nambu A, Tsujimoto T, Matsuzaki R, Kyuhou S, Gemba H. Studies on integrative functions of the human frontal association cortex with MEG. Brain Res Cogn Brain Res 1 996; 5:165-74.
Schalk G. Personal communication. 2005.
Schalk G. Effective brain-computer interfacing using BCI2000. Conf Proc IEEE Eng Med Biol Soc 2009;2009:5498-501.
Schalk G, McFarland DJ, Hinterberger T, Birbaumer N, Wolpaw JR. BCI2000: a general-purpose brain-computer interface. IEEE Trans Biomed Eng 2004; 5 1 : 1034-43.
Schalk G, Kubánek J, Miller KJ, Anderson NR, Leuthardt EC, Ojemann JG, Limbrick D, Moran D, Gerhardt LA, Wolpaw JR. Decoding two-dimensional movement trajectories using electrocorticographic signals in humans. J Neural Eng 2007; 4:264-75.
Schalk G, Brunner P, Gerhardt LA, Bischof H, Wolpaw JR. Brain-computer interfaces (BCIs): detection instead of classification. J Neurosci Methods 2008a; 167:51-62.
Schalk G, Leuthardt EC, Brunner P, Ojemann JG, Gerhardt LA, Wolpaw JR. Real-time detection of event-related brain activity. Neuroimage 2008b; 43:245-49.
Schwartz AB, Kettner RE, Georgopoulos AP. Primate motor cortex and free arm movements to visual targets in three-dimensional space. I. Relations between single cell discharge and direction of movement. J Neurosci 1988; 8:2913-27.
Schwartz AB, Moran DW, Reina GA. Differential representation of perception and action in the frontal cortex. Science 2004; 303:380-83.
Schwartz AB, Cui XT, Weber DJ, Moran DW. Brain-controlled interfaces: movement restoration with neural prosthetics. Neuron 2006; 52:205-20.
Vetter RG, Williams JC, Hetke JF, Nunamaker EA, Kipke DR. Chronic neural recording using silicon-substrate microelectrode arrays implanted in cerebral cortex. IEEE Trans Biomed Eng 2004;51:896-904.
Vidal JJ. Toward direct brain-computer communication. Annu Rev Biophys Bioeng 1973; 2:157-80.
Wisneski KJ, Anderson N, Schalk G, Smyth M, Moran D, Leuthardt EC. Unique cortical physiology associated with ipsilateral hand movements and neuroprosthetic implications. Stroke 2008;39:3351-59.
Wolpaw JR, Birbaumer N. Brain-computer interfaces for communication and control. In: Selzer M, Clarke S, Cohen L. Duncan P, Gage F (Editors). Textbook of Neural Repair and Rehabilitation: Neural Repair and Plasticity: Volume 1. New York: Cambridge University Press; 2006, p. 602-14.
Wolpaw JR, Birbaumer N, McFarland DJ, Pfurtschellar G, Vaughan TM. Brain-computer interfaces for communication and control. Clin Neurophysiol 2002; 1 13:767-91.
Wolpaw JR, McFarland DJ. Control of a two-dimensional movement signal by a non-invasive brain-computer interface in humans. Presented at the Society for Neuroscience 2004; San Diego, California.
Wolpaw JR, McFarland DJ, Neat GW, Forneris CA. An EEG-based brain-computer interface for cursor control. Electroencephalogr Clin Neurophys 1991; 78:252-59.
Wolpaw JR, McFarland DJ, Vaughan TM, Schalk G. The Wadsworth Center brain-computer interface (BCI) research and development program. IEEE Trans Neural Syst Rehabil Eng 2003; 11:204-07.
Zimmerman JE, Theine P, Harding JT. Design and operation of stable rf-biased superconducting point-contact quantum devices, etc. J App Phys 1 970; 41:1 572-80.
Nicholas R. Anderson, Ph.D.1 and Elise M. DeVries, BS2
1 San Diego, California
2 Chicago, Illinois