Towards human BCI applications based on cognitive brain systems: an investigation of neural signals recorded from the dorsolateral prefrontal cortex.

One of the critical issues in brain-computer interface (BCI) research is how to translate a person's intention into brain signals for controlling computer programs. The motor system is currently the primary focus, where signals are obtained during imagined motor responses. However, cognitive brain systems are also attractive candidates, in that they may be more amenable to conscious control, yielding better regulation of magnitude and duration of localized brain activity. We report on a proof of principle study for the potential use of a higher cognitive system for BCI, namely the working memory (WM) system. We show that mental calculation reliably activates the WM network as measured with functional magnetic resonance imaging (fMRI). Moreover, activity in the dorsolateral prefrontal cortex (DLPFC) indicates that this region is active for the duration of mental processing. This supports the notion that DLPFC can be activated, and remains active, at will. Further confirmation is obtained from a patient with an implanted electrode grid for diagnostic purposes, in that gamma power within DLPFC increases during mental calculation and remains elevated for the duration thereof. These results indicate that cortical regions involved in higher cognitive functions may serve as a readily self-controllable input for BCI applications. It also shows that fMRI is an effective tool for identifying function-specific foci in individual subjects for subsequent placement of cortical electrodes. The fact that electrocorticographic (ECoG) signal confirmed the functional localization of fMRI provides a strong argument for incorporating fMRI in BCI research.     

Phase synchronization for the recognition of mental tasks in a brain-computer interface

Brain-computer interfaces (BCIs) may be a future communication channel for motor-disabled people. In surface electroencephalogram (EEG)-based BCIs, the extracted features are often derived from spectral estimates and autoregressive models. We examined the usefulness of synchronization between EEG signals for classifying mental tasks. To this end, we investigated the performance of features derived from the phase locking value (PLV) and from the spectral coherence and compared them to the classification rates resulting from the power densities in alpha, beta1, beta2, and 8-30-Hz frequency bands. Five recordings of 60 min, acquired from three subjects while performing three different mental tasks, were analyzed offline. No artifacts were removed or rejected. We noticed significant differences between PLV and mean spectral coherence. For sole use of synchronization measures, classification accuracies up to 62% were achieved. In general, the best result was obtained combining phase synchronization measures with alpha power spectral density estimates. The results demonstrate that phase synchronization provides relevant information for the classification of spontaneous EEG during mental tasks.     

Comparison of linear, nonlinear, and feature selection methods for EEG signal classification

The reliable operation of brain-computer interfaces (BCIs) based on spontaneous electroencephalogram (EEG) signals requires accurate classification of multichannel EEG. The design of EEG representations and classifiers for BCI are open research questions whose difficulty stems from the need to extract complex spatial and temporal patterns from noisy multidimensional time series obtained from EEG measurements. The high-dimensional and noisy nature of EEG may limit the advantage of nonlinear classification methods over linear ones. This paper reports the results of a linear (linear discriminant analysis) and two nonlinear classifiers (neural networks and support vector machines) applied to the classification of spontaneous EEG during five mental tasks, showing that nonlinear classifiers produce only slightly better classification results. An approach to feature selection based on genetic algorithms is also presented with preliminary results of application to EEG during finger movement.     

Using nonlinear dynamic metric tools for characterizing brainstructures

The collective dynamic behavior of the neural mass of different brain structures can be assessed from electroencephalographic recordings with depth electrodes measurements at regular time intervals (EEG time series). In recent years, the cheery of nonlinear dynamics has developed methods for quantitative analysis of experimental time series. The aim of this article is to report a new attempt to characterize global brain dynamics through electrical activity using these nonlinear dynamical metric tools. In addition, the authors study the dependence of the metric magnitudes on brain structure. The methods employed in this work are independent of any modeling of brain activity. They rely solely on the analysis of data obtained from a single variable time series. The authors analyze the EEG signals from depth electrodes that intersect different brain anatomical structures in a patient with refractory epilepsy prone to surgical treatment. The electrical signal provided by this type of electrode guarantees a low noise signal     

The impact of loss of control on movement BCIs

Brain-computer interfaces (BCIs) are known to suffer from spontaneous changes in the brain activity. If changes in the mental state of the user are reflected in the brain signals used for control, the behavior of a BCI is directly influenced by these states. We investigate the influence of a state of loss of control in a variant of Pacman on the performance of BCIs based on motor control. To study the effect a temporal loss of control has on the BCI performance, BCI classifiers were trained on electroencephalography (EEG) recorded during the normal control condition, and the classification performance on segments of EEG from the normal and loss of control condition was compared. Classifiers based on event-related desynchronization unexpectedly performed significantly better during the loss of control condition; for the event-related potential classifiers there was no significant difference in performance.     

Application of tripolar concentric electrodes and prefeature selection algorithm for brain-computer interface.

For persons with severe disabilities, a brain-computer interface (BCI) may be a viable means of communication. Lapalacian electroencephalogram (EEG) has been shown to improve classification in EEG recognition. In this work, the effectiveness of signals from tripolar concentric electrodes and disc electrodes were compared for use as a BCI. Two sets of left/right hand motor imagery EEG signals were acquired. An autoregressive (AR) model was developed for feature extraction with a Mahalanobis distance based linear classifier for classification. An exhaust selection algorithm was employed to analyze three factors before feature extraction. The factors analyzed were 1) length of data in each trial to be used, 2) start position of data, and 3) the order of the AR model. The results showed that tripolar concentric electrodes generated significantly higher classification accuracy than disc electrodes.     

Application of Tripolar Concentric Electrodes and Prefeature Selection Algorithm for Brain–Computer Interface

For persons with severe disabilities, a brain-computer interface (BCI) may be a viable means of communication. Lapalacian electroencephalogram (EEG) has been shown to improve classification in EEG recognition. In this work, the effectiveness of signals from tripolar concentric electrodes and disc electrodes were compared for use as a BCI. Two sets of left/right hand motor imagery EEG signals were acquired. An autoregressive (AR) model was developed for feature extraction with a Mahalanobis distance based linear classifier for classification. An exhaust selection algorithm was employed to analyze three factors before feature extraction. The factors analyzed were 1) length of data in each trial to be used, 2) start position of data, and 3) the order of the AR model. The results showed that tripolar concentric electrodes generated significantly higher classification accuracy than disc electrodes.     

The use of EEG modifications due to motor imagery for brain-computer interfaces

The opening of a communication channel between brain and computer [brain-computer interface (BCI)] is possible by using changes in electroencephalogram (EEG) power spectra related to the imagination of movements. In this paper, we present results obtained by recording EEG during an upper limb motor imagery task in a total of 18 subjects by using low-resolution surface Laplacian, different linear and quadratic classifiers, as well as a variable number of scalp electrodes, from 2 to 26. The results (variable correct classification rate of mental imagery between 75% and 95%) suggest that it is possible to recognize quite reliably ongoing mental movement imagery for BCI applications.     

EEG changes accompanying learned regulation of 12-Hz EEG activity

We analyzed 15 sessions of 64-channel electroencephalographic (EEG) data recorded from a highly trained subject during sessions in which he attempted to regulate power at 12 Hz over his left- and right-central scalp to control the altitude of a cursor moving toward target boxes placed at the top-, middle-, or bottom-right of a computer screen. We used infomax independent component analysis (ICA) to decompose 64-channel EEG data from trials in which the subject successfully up- or down-regulated the measured EEG signals. Applying time-frequency analysis to the time courses of activity of several of the resulting 64 independent EEG components revealed that successful regulation of the measured activity was accompanied by extensive, asymmetrical changes in power and coherence, at both nearby and distant frequencies, in several parts of cortex. A more complete understanding of these phenomena could help to explain the nature and locus of learned regulation of EEG rhythms and might also suggest ways to further optimize the performance of brain-computer interfaces.     

生理性精神疲劳EEG信号分析技术

从脑电（EEG）信号分析的角度出发,探讨了传统的生理性精神疲劳的分析方法和现阶段基于EEG信号分析的生理性精神疲劳所采用的新技术、新方法,构建了生理性精神疲劳的分析系统,展望了生理性精神疲劳EEG信号分析技术的发展趋势。     

Utilizing gamma band to improve mental task based brain-computer interface design.

A common method for designing brain-computer Interface (BCI) is to use electroencephalogram (EEG) signals extracted during mental tasks. In these BCI designs, features from EEG such as power and asymmetry ratios from delta, theta, alpha, and beta bands have been used in classifying different mental tasks. In this paper, the performance of the mental task based BCI design is improved by using spectral power and asymmetry ratios from gamma (24-37 Hz) band in addition to the lower frequency bands. In the experimental study, EEG signals extracted during five mental tasks from four subjects were used. Elman neural network (ENN) trained by the resilient backpropagation algorithm was used to classify the power and asymmetry ratios from EEG into different combinations of two mental tasks. The results indicated that ((1) the classification performance and training time of the BCI design were improved through the use of additional gamma band features; (2) classification performances were nearly invariant to the number of ENN hidden units or feature extraction method.     

Grael脑电放大器与深度学习的手势实时识别研究

手势识别是人机交互的关键。为了能够更好地实现脑电信号与肌电信号的融合,精准地识别人体的运动,本文建立了一套基于Grael脑电放大器的手势动作实时检测识别的研究系统。通过Grael脑电放大器和Curry8系统采集5个通道的8种不同手势的表面肌电信号(sEMG),并对采集到的sEMG信号进行滤波去噪、滑动窗口分割以及特征提取等预处理的操作;最后采用几种常用的分类器与卷积神经网络(CNN)对不同手势的sEMG信号进行实时分类识别。结果表明CNN的识别准确率最高,能达到92.98%;对每个手势动作进行30次实时识别检测,结果显示识别延迟大概在1～1.5 s,实时识别的精度可高达90%。该系统为将来研究脑电信号与肌电信号的融合提供了一个可行的方法,在人机交互方面展现了巨大的潜力和应用空间。     

A time-frequency approach for newborn seizure detection

Techniques previously designed for seizure detection in newborns using the electroencephalogram (EEG) have been relatively inefficient due to their assumption of local stationarity of the EEG. To overcome the problem raised by the nonstationarity of the EEG signal, current methods are extended to a time-frequency approach. This allows the analysis and characterization of the different newborn EEG patterns that are intended to be the first step toward an automatic time-frequency seizure detection and classification. An in-depth analysis of both the autocorrelation and spectrum seizure detection techniques identified the detection criteria that can be extended to the time-frequency domain. The selected method uses a high-resolution reduced interference time-frequency distribution referred to as the B-distribution (BD). Here, the authors present the various patterns of observed time-frequency seizure signals and relate them to current knowledge of seizures. In particular, initial results indicate that a quasilinear instantaneous frequency (IF) can be used as a critical feature of the EEG seizure characteristics     

Low-resolution electromagnetic tomography neurofeedback

Through continuous feedback of the electroencephalogram (EEG) humans can learn how to shape their brain electrical activity in a desired direction. The technique is known as EEG biofeedback, or neurofeedback, and has been used since the late 1960s in research and clinical applications. A major limitation of neurofeedback relates to the limited information provided by a single or small number of electrodes placed on the scalp. We establish a method for extracting and feeding back intracranial current density and we carry out an experimental study to ascertain the ability of the participants to drive their own EEG power in a desired direction. To derive current density within the brain volume, we used the low-resolution electromagnetic tomography (LORETA). Six undergraduate students (three males, three females) underwent tomographic neurofeedback (based on 19 electrodes placed according to the 10-20 system) to enhance the current density power ratio between the frequency bands beta (16-20 Hz) and alpha (8-10 Hz). According to LORETA modeling, the region of interest corresponded to the Anterior Cingulate (cognitive division). The protocol was designed to improve the performance of the subjects on the dimension of sustained attention. Two hypotheses were tested: 1) that the beta/alpha current density power ratio increased over sessions and 2) that by the end of the training subjects acquired the ability of increasing that ratio at will. Both hypotheses received substantial experimental support in this study. This is the first application of an EEG inverse solution to neurofeedback. Possible applications of the technique include the treatment of epileptic foci, the rehabilitation of specific brain regions damaged as a consequence of traumatic brain injury and, in general, the training of any spatial specific cortical electrical activity. These findings may also have relevant consequences for the development of brain-computer interfaces.     

Current trends in Graz Brain-Computer Interface (BCI) research.

This paper describes a research approach to develop a brain-computer interface (BCI) based on recognition of subject-specific EEG patterns. EEG signals recorded from sensorimotor areas during mental imagination of specific movements are classified on-line and used e.g. for cursor control. In a number of on-line experiments, various methods for EEG feature extraction and classification have been evaluated.     

A natural basis for efficient brain-actuated control.

The prospect of noninvasive brain-actuated control of computerized screen displays or locomotive devices is of interest to many and of crucial importance to a few 'locked-in' subjects who experience near total motor paralysis while retaining sensory and mental faculties. Currently several groups are attempting to achieve brain-actuated control of screen displays using operant conditioning of particular features of the spontaneous scalp electroencephalogram (EEG) including central mu-rhythms (9-12 Hz). A new EEG decomposition technique, independent component analysis (ICA), appears to be a foundation for new research in the design of systems for detection and operant control of endogenous EEG rhythms to achieve flexible EEG-based communication. ICA separates multichannel EEG data into spatially static and temporally independent components including separate components accounting for posterior alpha rhythms and central mu activities. We demonstrate using data from a visual selective attention task that ICA-derived mu-components can show much stronger spectral reactivity to motor events than activity measures for single scalp channels. ICA decompositions of spontaneous EEG would thus appear to form a natural basis for operant conditioning to achieve efficient and multidimensional brain-actuated control in motor-limited and locked-in subjects.     

Specific monitoring of neonatal brain function with optimizedfrequency bands

Early detection of altered brain function can be helpful in preventing the development of serious brain damage. Power spectral analysis of continuous EEG is an established tool in corresponding clinical monitoring. The commonly used EEG power classifiers are based on the power within particular frequency bands. It can be supposed that individually adapted frequency bands allow a more specific monitoring of altered brain function than the commonly used standard EEG frequency bands. In order to test this hypothesis and provide an appropriate signals analysis approach. A hybrid analysis system (HAS) containing variable frequency band power estimators and artificial neural networks (ANNs) was trained with regard to brain function during well-defined states of hemorrhagic hypotension. The aim of the methodical study presented in this article was to investigate whether specific EEG frequency band power parameters can be found by means of this HAS, which enables better detection and classification of moderately reduced brain supply than the classical fixed frequency bands     

Decoding intra-limb and inter-limb kinematics during treadmill walking from scalp electroencephalographic (EEG) signals

Brain-machine interface (BMI) research has largely been focused on the upper limb. Although restoration of gait function has been a long-standing focus of rehabilitation research, surprisingly very little has been done to decode the cortical neural networks involved in the guidance and control of bipedal locomotion. A notable exception is the work by Nicolelis' group at Duke University that decoded gait kinematics from chronic recordings from ensembles of neurons in primary sensorimotor areas in rhesus monkeys. Recently, we showed that gait kinematics from the ankle, knee, and hip joints during human treadmill walking can be inferred from the electroencephalogram (EEG) with decoding accuracies comparable to those using intracortical recordings. Here we show that both intra- and inter-limb kinematics from human treadmill walking can be achieved with high accuracy from as few as 12 electrodes using scalp EEG. Interestingly, forward and backward predictors from EEG signals lagging or leading the kinematics, respectively, showed different spatial distributions suggesting distinct neural networks for feedforward and feedback control of gait. Of interest is that average decoding accuracy across subjects and decoding modes was ~0.68±0.08, supporting the feasibility of EEG-based BMI systems for restoration of walking in patients with paralysis.     

A data-reduction process for long-term EEGs. Feature extractionthrough digital processing in a multiresolution framework

Describes a contribution to a data-reduction process to be used with long-term EEGs. Since typical long-term EECs recorded from depth electrodes are extended over several days, while epilepsy may be characterized by occasional transients, data reduction is an important consideration for the electroencephalographer. The electroencephalographer detects epileptic activity by visual inspection of the EEG, which is a time-consuming procedure for records that are days long. The result obtained with the authors' proposed algorithm is the selection of segments of EEG where a transient is detected; then these segments are reviewed by an expert. The authors' primary objective is to minimize the visual inspection process, presenting to the clinician only selected segments of the EEG. Throughout this article, no distinction is made among the wide variety of epileptiform transients. The only objective of the algorithm is to separate background activity from epileptiform activity     

Brain-computer interfaces for 1-D and 2-D cursor control: designs using volitional control of the EEG spectrum or steady-state visual evoked potentials.

We have developed and tested two electroencephalogram (EEG)-based brain-computer interfaces (BCI) for users to control a cursor on a computer display. Our system uses an adaptive algorithm, based on kernel partial least squares classification (KPLS), to associate patterns in multichannel EEG frequency spectra with cursor controls. Our first BCI, Target Practice, is a system for one-dimensional device control, in which participants use biofeedback to learn voluntary control of their EEG spectra. Target Practice uses a KPLS classifier to map power spectra of 62-electrode EEG signals to rightward or leftward position of a moving cursor on a computer display. Three subjects learned to control motion of a cursor on a video display in multiple blocks of 60 trials over periods of up to six weeks. The best subject's average skill in correct selection of the cursor direction grew from 58% to 88% after 13 training sessions. Target Practice also implements online control of two artifact sources: 1) removal of ocular artifact by linear subtraction of wavelet-smoothed vertical and horizontal electrooculograms (EOG) signals, 2) control of muscle artifact by inhibition of BCI training during periods of relatively high power in the 40-64 Hz band. The second BCI, Think Pointer, is a system for two-dimensional cursor control. Steady-state visual evoked potentials (SSVEP) are triggered by four flickering checkerboard stimuli located in narrow strips at each edge of the display. The user attends to one of the four beacons to initiate motion in the desired direction. The SSVEP signals are recorded from 12 electrodes located over the occipital region. A KPLS classifier is individually calibrated to map multichannel frequency bands of the SSVEP signals to right-left or up-down motion of a cursor on a computer display. The display stops moving when the user attends to a central fixation point. As for Target Practice, Think Pointer also implements wavelet-based online removal of ocular artifact; however, in Think Pointer muscle artifact is controlled via adaptive normalization of the SSVEP. Training of the classifier requires about 3 min. We have tested our system in real-time operation in three human subjects. Across subjects and sessions, control accuracy ranged from 80% to 100% correct with lags of 1-5 s for movement initiation and turning. We have also developed a realistic demonstration of our system for control of a moving map display (http://ti.arc.nasa.gov/).