The quantitative extraction and topographic mapping of the abnormal components in the clinical EEG

Evaluating User Experience of Actual and Imagined Movement in BCI Gaming

Most research on Brain-Computer Interfaces (BCI) focuses on developing ways of expression for disabled people who are not able to communicate through other means. Recently it has been shown that BCI can also be used in games to give users a richer experience and new ways to interact with a computer or game console. This paper describes research conducted to find out what the differences are between using actual and imagined movement as modalities in a BCI game. Results show that there are significant differences in user experience and that actual movement is a more robust way of communicating through a BCI.

Model based generalization analysis of common spatial pattern in brain computer interfaces

In the motor imagery based Brain Computer Interface (BCI) research, Common Spatial Pattern (CSP) algorithm is used widely as a spatial filter on multi-channel electroencephalogram (EEG) recordings. Recently the overfitting effect of CSP has been gradually noticed, but what influence the overfitting is still unclear. In this work, the generalization of CSP is investigated by a simple linear mixing model. Several factors in this model are discussed, and the simulation results indicate that channel numbers and the correlation between signals influence the generalization of CSP significantly. A larger number of training trials and a longer time length of the trial would prevent overfitting. The experiments on real data also verify our conclusion.

Robust common spatial filters with a maxmin approach

Electroencephalographic signals are known to be non-stationary and easily affected by artifacts, therefore their analysis requires methods that can deal with noise. In this work we present two ways of calculating robust common spatial patterns under a maxmin approach. The worst-case objective function is optimized within prefixed sets of the covariance matrices that are defined either very simply as identity matrices or in a data driven way using PCA. We test common spatial filters derived with these two approaches with real world brain-computer interface (BCI) data sets in which we expect substantial "day-to-day" fluctuations (session transfer problem). We compare our results with the classical common spatial filters and show that both can improve the performance of the latter.

Finding stationary brain sources in EEG data

Neurophysiological measurements obtained from e.g. EEG or fMRI are inherently non-stationary because the properties of the underlying brain processes vary over time. For example, in Brain-Computer-Interfacing (BCI), deteriorating performance (bitrate) is a common phenomenon since the parameters determined during the calibration phase can be suboptimal under the application regime, where the brain state is different, e.g. due to increased tiredness or changes in the experimental paradigm. We show that Stationary Subspace Analysis (SSA), a time series analysis method, can be used to identify the underlying stationary and non-stationary brain sources from high-dimensional EEG measurements. Restricting the BCI to the stationary sources found by SSA can significantly increase the performance. Moreover, SSA yields topographic maps corresponding to stationary- and non-stationary brain sources which reveal their spatial characteristics.

Finding stationary subspaces in multivariate time series

Identifying temporally invariant components in complex multivariate time series is key to understanding the underlying dynamical system and predict its future behavior. In this Letter, we propose a novel technique, stationary subspace analysis (SSA), that decomposes a multivariate time series into its stationary and nonstationary part. The method is based on two assumptions: (a) the observed signals are linear superpositions of stationary and nonstationary sources; and (b) the nonstationarity is measurable in the first two moments. We characterize theoretical and practical properties of SSA and study it in simulations and cortical signals measured by electroencephalography. Here, SSA succeeds in finding stationary components that lead to a significantly improved prediction accuracy and meaningful topographic maps which contribute to a better understanding of the underlying nonstationary brain processes.

Improving BCI performance by task-related trial pruning

Noise in electroencephalography data (EEG) is an ubiquitous problem that limits the performance of brain computer interfaces (BCI). While typical EEG artifacts are usually removed by trial rejection or by filtering, noise induced in the data by the subject's failure to produce the required mental state is very harmful. Such "noise" effects are rather common, especially for naive subjects in their training phase and, thus, standard artifact removal methods would inevitably fail. In this paper, we present a novel method which aims to detect such defected trials taking into account the intended task by use of Relevant Dimensionality Estimation (RDE), a new machine learning method for denoising in feature space. In this manner, our method effectively "cleans" the training data and thus allows better BCI classification. Preliminary results conducted on a data set of 43 naive subjects show a significant improvement for 74% of the subjects.

The brain-computer interface cycle

Brain-computer interfaces (BCIs) have attracted much attention recently, triggered by new scientific progress in understanding brain function and by impressive applications. The aim of this review is to give an overview of the various steps in the BCI cycle, i.e., the loop from the measurement of brain activity, classification of data, feedback to the subject and the effect of feedback on brain activity. In this article we will review the critical steps of the BCI cycle, the present issues and state-of-the-art results. Moreover, we will develop a vision on how recently obtained results may contribute to new insights in neurocognition and, in particular, in the neural representation of perceived stimuli, intended actions and emotions. Now is the right time to explore what can be gained by embracing real-time, online BCI and by adding it to the set of experimental tools already available to the cognitive neuroscientist. We close by pointing out some unresolved issues and present our view on how BCI could become an important new tool for probing human cognition.

Extraction of percept-related induced local field potential during spontaneously reversing perception

The question of how perception arises from neuronal activity in the visual cortex is of fundamental importance in cognitive neuroscience. To address this question, we adopt a unique experimental paradigm in which bistable structure-from-motion (SFM) stimuli are employed to dissociate the visual input from perception while monitoring the cortical neural activity called local field potential (LFP). Consequently, the stimulus-evoked activity of LFP is not related to perception but the oscillatory induced activity of LFP may be percept-related. In this paper we focus on extracting the percept-related features of the induced activity from LFP in a monkey's visual cortex for decoding its bistable structure-from-motion perception. We first estimate the stimulus-evoked activity via a wavelet-based method and remove it from the single-trial LFP. We then use the common spatial patterns (CSP) approach to design spatial filters to extract the percept-related features from the remaining induced activity. We exploit the linear discriminant analysis (LDA) classifier on the extracted features to decode the reported perception on a single-trial basis. We demonstrate that our approach has excellent performance in estimating the stimulus-evoked activity, outperforming the Wiener filter, least mean square (LMS), and a local regression method called the locally weighted scatterplot smoothing (LOWESS), and that our approach is effective in extracting the discriminative features of the percept-related induced activity from LFP, which leads to excellent decoding performance. We also discover that the enhanced gamma band synchronization and reduced alpha band desynchronization may be the underpinnings of the induced activity.

A robust and self-paced BCI system based on a four class SSVEP paradigm: algorithms and protocols for a high-transfer-rate direct brain communication

In this paper, we present, with particular focus on the adopted processing and identification chain and protocol-related solutions, a whole self-paced brain-computer interface system based on a 4-class steady-state visual evoked potentials (SSVEPs) paradigm. The proposed system incorporates an automated spatial filtering technique centred on the common spatial patterns (CSPs) method, an autoscaled and effective signal features extraction which is used for providing an unsupervised biofeedback, and a robust self-paced classifier based on the discriminant analysis theory. The adopted operating protocol is structured in a screening, training, and testing phase aimed at collecting user-specific information regarding best stimulation frequencies, optimal sources identification, and overall system processing chain calibration in only a few minutes. The system, validated on 11 healthy/pathologic subjects, has proven to be reliable in terms of achievable communication speed (up to 70 bit/min) and very robust to false positive identifications.

BrainNetVis: analysis and visualization of brain functional networks

BrainNetVis is an application, written in Java, that displays and analyzes synchronization networks from brain signals. The program implements a number of network indices and visualization techniques. We demonstrate its use through a case study of left hand and foot motor imagery. The data sets were provided by the Berlin BCI group. Using this program we managed to find differences between the average left hand and foot synchronization networks by comparing them with the average idle state synchronization network.

Towards zero training for brain-computer interfacing

Electroencephalogram (EEG) signals are highly subject-specific and vary considerably even between recording sessions of the same user within the same experimental paradigm. This challenges a stable operation of Brain-Computer Interface (BCI) systems. The classical approach is to train users by neurofeedback to produce fixed stereotypical patterns of brain activity. In the machine learning approach, a widely adapted method for dealing with those variances is to record a so called calibration measurement on the beginning of each session in order to optimize spatial filters and classifiers specifically for each subject and each day. This adaptation of the system to the individual brain signature of each user relieves from the need of extensive user training. In this paper we suggest a new method that overcomes the requirement of these time-consuming calibration recordings for long-term BCI users. The method takes advantage of knowledge collected in previous sessions: By a novel technique, prototypical spatial filters are determined which have better generalization properties compared to single-session filters. In particular, they can be used in follow-up sessions without the need to recalibrate the system. This way the calibration periods can be dramatically shortened or even completely omitted for these 'experienced' BCI users. The feasibility of our novel approach is demonstrated with a series of online BCI experiments. Although performed without any calibration measurement at all, no loss of classification performance was observed.

Single-Trial Classification of Bistable Perception by Integrating Empirical Mode Decomposition, Clustering, and Support Vector Machine

We propose an empirical mode decomposition (EMD-) based method to extract features from the multichannel recordings of local field potential (LFP), collected from the middle temporal (MT) visual cortex in a macaque monkey, for decoding its bistable structure-from-motion (SFM) perception. The feature extraction approach consists of three stages. First, we employ EMD to decompose nonstationary single-trial time series into narrowband components called intrinsic mode functions (IMFs) with time scales dependent on the data. Second, we adopt unsupervised K-means clustering to group the IMFs and residues into several clusters across all trials and channels. Third, we use the supervised common spatial patterns (CSP) approach to design spatial filters for the clustered spatiotemporal signals. We exploit the support vector machine (SVM) classifier on the extracted features to decode the reported perception on a single-trial basis. We demonstrate that the CSP feature of the cluster in the gamma frequency band outperforms the features in other frequency bands and leads to the best decoding performance. We also show that the EMD-based feature extraction can be useful for evoked potential estimation. Our proposed feature extraction approach may have potential for many applications involving nonstationary multivariable time series such as brain-computer interfaces (BCI).

An MEG-based brain-computer interface (BCI)

Brain-computer interfaces (BCIs) allow for communicating intentions by mere brain activity, not involving muscles. Thus, BCIs may offer patients who have lost all voluntary muscle control the only possible way to communicate. Many recent studies have demonstrated that BCIs based on electroencephalography (EEG) can allow healthy and severely paralyzed individuals to communicate. While this approach is safe and inexpensive, communication is slow. Magnetoencephalography (MEG) provides signals with higher spatiotemporal resolution than EEG and could thus be used to explore whether these improved signal properties translate into increased BCI communication speed. In this study, we investigated the utility of an MEG-based BCI that uses voluntary amplitude modulation of sensorimotor mu and beta rhythms. To increase the signal-to-noise ratio, we present a simple spatial filtering method that takes the geometric properties of signal propagation in MEG into account, and we present methods that can process artifacts specifically encountered in an MEG-based BCI. Exemplarily, six participants were successfully trained to communicate binary decisions by imagery of limb movements using a feedback paradigm. Participants achieved significant mu rhythm self control within 32 min of feedback training. For a subgroup of three participants, we localized the origin of the amplitude modulated signal to the motor cortex. Our results suggest that an MEG-based BCI is feasible and efficient in terms of user training.

Spatio-temporal decomposition of the electroencephalogram in patients with cirrhosis

BACKGROUND/AIMS

Slowing of the electroencephalogram (EEG) is a recognised feature of hepatic encephalopathy but its diagnostic sensitivity is indeterminate. Recent advances in EEG analysis should provide better quantifiable/more informative data. The aim of this study was to isolate and determine the scalp distribution of the posterior basic rhythm, in patients with cirrhosis, using a technique for spatio-temporal decomposition (SEDACA) of the EEG.

METHODS

One hundred and ten patients with cirrhosis, classified, using clinical and psychometric criteria, as neuropsychiatrically unimpaired or as having minimal/overt hepatic encephalopathy were studied. Eyes-closed, awake EEGs were obtained and subjected to standard spectral analysis and spatio-temporal decomposition. Control data were obtained from 26 reference EEGs.

RESULTS

The error in the estimate of the SEDACA-derived mean dominant frequency was lower than for the standard EEG derivation (P<0.00001). The SEDACA-derived spectral estimates correlated better with neuropsychiatric status and allowed differentiation of the patients with minimal hepatic encephalopathy from the reference population. The SEDACA-derived spatial information showed an anteriorization of the posterior basic rhythm, which became more prominent as the degree of neuropsychiatric impairment increased (P=0.00052).

CONCLUSIONS

Analysis of the EEG utilising SEDACA provides significantly more diagnostic information on the neuropsychiatric status of patients with cirrhosis than obtained conventionally.

Classification of movement intention by spatially filtered electromagnetic inverse solutions

We couple standardized low-resolution electromagnetic tomography, an inverse solution for electroencephalography (EEG) and the common spatial pattern, which is here conceived as a data-driven beamformer, to classify the benchmark BCI (brain-computer interface) competition 2003, data set IV. The data set is from an experiment where a subject performed a self-paced left and right finger tapping task. Available for analysis are 314 training trials whereas 100 unlabelled test trials have to be classified. The EEG data from 28 electrodes comprise the recording of the 500 ms before the actual finger movements, hence represent uniquely the left and right finger movement intention. Despite our use of an untrained classifier, and our extraction of only one attribute per class, our method yields accuracy similar to the winners of the competition for this data set. The distinct advantages of the approach presented here are the use of an untrained classifier and the processing speed, which make the method suitable for actual BCI applications. The proposed method is favourable over existing classification methods based on an EEG inverse solution, which rely either on iterative algorithms for single-trial independent component analysis or on trained classifiers.

Recipes for the linear analysis of EEG

In this paper, we describe a simple set of "recipes" for the analysis of high spatial density EEG. We focus on a linear integration of multiple channels for extracting individual components without making any spatial or anatomical modeling assumptions, instead requiring particular statistical properties such as maximum difference, maximum power, or statistical independence. We demonstrate how corresponding algorithms, for example, linear discriminant analysis, principal component analysis and independent component analysis, can be used to remove eye-motion artifacts, extract strong evoked responses, and decompose temporally overlapping components. The general approach is shown to be consistent with the underlying physics of EEG, which specifies a linear mixing model of the underlying neural and non-neural current sources.

Principal component analysis for reduction of ocular artefacts in event-related potentials of normal and dyslexic children

OBJECTIVE

The aim of this study was to reduce ocular artefacts in single trial event-related potentials (ERPs) recorded in normal and in dyslexic children.

METHODS

ERPs were recorded during passive and active reading of centrally presented alphabetic letters and non alphabetic symbols. EEG was recorded from 10 EEG locations using the 10-20 system. Diagonal EOG from the right eye was also recorded. Principal component analysis (PCA) was applied in order to reduce ocular artefacts: the first or the second principal component (PC) was subtracted when the correlation coefficient between the component and EOG was greater or equal to 0.9 or 0.95, respectively. Performance of the method was tested on simulated and real data, on both single and averaged trials, varying EOG amplitude and artefact transmission characteristics.

RESULTS

Applying the method to real recordings from normal and dyslexic children, we obtained a significant increase in the number of useful trials. In normal children we retrieved 41.0% of the rejected trials in passive and 39.1% in active reading. In dyslexic children 36.7 and 32.2% of the rejected trials in passive and active reading could be included in the respective averages.

CONCLUSIONS

The method allows an increase in the number of trials suitable for averaging, a great improvement in ERP quality and a reduction in the recording time.

A novel spatio-temporal decomposition of the EEG: derivation, validation and clinical application

OBJECTIVE

To obtain clinically useful graphical and numerical data on the distribution of activities in the EEG using a novel type of spatio-temporal decomposition.

METHODS

The EEG is divided into 1/4 s epochs. An approximation to the spatial distribution of the locally dominant activity in each epoch is represented as a point in a spatial component space. Points representing epochs dominated by activity from the same source form a cluster. The centres of these clusters represent the global spatial component of each source. As each spatial component is identified, its corresponding temporal activity is removed from the record, allowing activity from sources with smaller amplitude to become dominant in the reduced record. Successive components are identified in the reduced record. The method was applied to 40 normal EEGs and features were identified, which were common to them all. The method was also applied to 4 separate records with different forms of focal abnormality.

RESULTS

The method successfully separated components from the EEG representing alpha rhythm, eye artefact, electrode artefact and EEG. In 40 normal EEGs the method isolated spatial components that were common to all EEGs, and in 4 abnormal EEGs it achieved a high degree of mutual separation of alpha rhythm, focal spikes, focal theta and focal delta activities.

CONCLUSIONS

The method achieved a high degree of mutual separation of the EEG components and successfully differentiated the artefacts due to eye movement, ECG and electrode faults. The clinical implications are discussed.

Identification of the temporal components of seizure onset in the scalp EEG

BACKGROUND

The identification of the earliest indication of rhythmical oscillations and paroxysmal events associated with an epileptic seizure is paramount in identifying the location of the seizure onset in the scalp EEG. In this work, data-dependent filters are designed that can help reveal obscure activity at the onset of seizures in problematic EEGs.

METHODS

Data-dependent filters were designed using temporal patterns common to selected segments from pre-ictal and ictal portions of the scalp EEG. Temporal patterns that accounted for more variance in the ictal segment than in the pre-ictal segment of the scalp EEG were used to form the filters.

RESULTS

Application of the filters to the scalp EEG revealed temporal components in the seizure onset in the scalp recording that were not obvious in the unfiltered EEG. Examination of the filtered EEG enabled the onset of the seizure to be recognized earlier in the recording. The utility of the filters was confirmed qualitatively by comparing the scalp recording to the intracranial recording and quantitatively by calculating correlation coefficients between the scalp and intracranial recordings before and after filtering.

CONCLUSION

The data-dependent approach to EEG filter design allows automatic detection of the basic frequencies present in the seizure onset. This approach is more effective than narrow band-pass filtering for eliminating artifactual and other interference that can obscure the onset of a seizure. Therefore, temporal-pattern filtering facilitates the identification of seizure onsets in challenging scalp EEGs.

Optimal spatial filtering of single trial EEG during imagined hand movement

The development of an electroencephalograph (EEG)-based brain-computer interface (BCI) requires rapid and reliable discrimination of EEG patterns, e.g., associated with imaginary movement. One-sided hand movement imagination results in EEG changes located at contra- and ipsilateral central areas. We demonstrate that spatial filters for multichannel EEG effectively extract discriminatory information from two populations of single-trial EEG, recorded during left- and right-hand movement imagery. The best classification results for three subjects are 90.8%, 92.7%, and 99.7%. The spatial filters are estimated from a set of data by the method of common spatial patterns and reflect the specific activation of cortical areas. The method performs a weighting of the electrodes according to their importance for the classification task. The high recognition rates and computational simplicity make it a promising method for an EEG-based brain-computer interface.

Real-time EEG analysis with subject-specific spatial patterns for a brain-computer interface (BCI)

Electroencephalogram (EEG) recordings during right and left motor imagery allow one to establish a new communication channel for, e.g., patients with amyotrophic lateral sclerosis. Such an EEG-based brain-computer interface (BCI) can be used to develop a simple binary response for the control of a device. Three subjects participated in a series of on-line sessions to test if it is possible to use common spatial patterns to analyze EEG in real time in order to give feedback to the subjects. Furthermore, the classification accuracy that can be achieved after only three days of training was investigated. The patterns are estimated from a set of multichannel EEG data by the method of common spatial patterns and reflect the specific activation of cortical areas. By construction, common spatial patterns weight each electrode according to its importance to the discrimination task and suppress noise in individual channels by using correlations between neighboring electrodes. Experiments with three subjects resulted in an error rate of 2, 6 and 14% during on-line discrimination of left- and right-hand motor imagery after three days of training and make common spatial patterns a promising method for an EEG-based brain-computer interface.

Classification of movement-related EEG in a memorized delay task experiment

OBJECTIVES

We studied the activation of cortical motor areas during a memorized delay task with a classification technique.

METHODS

Multichannel EEG was recorded during the sequence of warning stimulus, visual cue, reaction stimulus, and actual execution of hand or foot movements. Two different approaches are presented: first, we trained a classifier on data from the time segments immediately preceding the actual movements, and analyzed the whole recordings in overlapping segments with this fixed classifier. The classification rates obtained as a function of experimental time reflect the activation of the same cortical areas that are active during the actual movements. In the second approach, we trained classifiers on data segments with the same latency in time as the data tested ('running classifiers'). By this, we checked whether we could detect event-related activity sufficiently marked to allow for correct classification.

RESULTS

With the fixed classifier approach we found two maxima of classification: one maximum after processing of the visual cue corresponding to an activation of motor cortex without overt movement, and a second maximum at the time of the actual movement. The first maximum relates to a very short-lived brain state, in the order of 300 ms, while the broad second maximum (1.5 s) indicates a very stable and long-lasting activation.

CONCLUSIONS

With the running classifier approach we found similar maxima as with the fixed classifier, indicating that only the activity of motor areas is relevant for classification. Possible implications of our findings for the development of a brain computer interface (BCI) are discussed.

Visually guided motor imagery activates sensorimotor areas in humans

Stimulus-related changes in ongoing electroencephalography (EEG) over sensorimotor areas were investigated during a visually cued motor imagery task. Four subjects were instructed to imagine one-sided hand movements in response to visual cue stimuli. The EEG was recorded from central areas using 27 electrodes set at distances of 2.5 cm. The method of common spatial filters was used to extract discriminatory information of EEG patterns recorded during the two motor imagery conditions. Single EEG trials were classified in intervals of 250 ms for a 8-s period starting 3 s prior to stimulus presentation. The results suggest that perception of the visual cue stimulus modifies oscillations in sensorimotor areas specific to the indicated hand starting as soon as 250-500 ms after stimulus onset.

Designing optimal spatial filters for single-trial EEG classification in a movement task

We devised spatial filters for multi-channel EEG that lead to signals which discriminate optimally between two conditions. We demonstrate the effectiveness of this method by classifying single-trial EEGs, recorded during preparation for movements of the left or right index finger or the right foot. The classification rates for 3 subjects were 94, 90 and 84%, respectively. The filters are estimated from a set of multichannel EEG data by the method of Common Spatial Patterns, and reflect the selective activation of cortical areas. By construction, we obtain an automatic weighting of electrodes according to their importance for the classification task. Computationally, this method is parallel by nature, and demands only the evaluation of scalar products. Therefore, it is well suited for on-line data processing. The recognition rates obtained with this relatively simple method are as good as, or higher than those obtained previously with other methods. The high recognition rates and the method's procedural and computational simplicity make it a particularly promising method for an EEG-based brain-computer interface.

Common spatial subspace decomposition applied to analysis of brain responses under multiple task conditions: a simulation study

A method, called common spatial subspace decomposition, is presented which can extract signal components specific to one condition from multiple magnetoencephalography/electroencephalography data sets of multiple task conditions. Signal matrices or covariance matrices are decomposed using spatial factors common to multiple conditions. The spatial factors and corresponding spatial filters are then dissociated into specific and common parts, according to the common spatial subspace which exists among the data sets. Finally, the specific signal components are extracted using the corresponding spatial filters and spatial factors. The relationship between this decomposition and spatio-temporal source models is described in this paper. Computer simulations suggest that this method can facilitate the analysis of brain responses under multiple task conditions and merits further application.

Trends in EEG source localization

The concepts underlying the quantitative localization of the sources of the EEG inside the brain are reviewed along with the current and emerging approaches to the problem. The concepts mentioned include monopolar and dipolar source models and head models ranging from the spherical to the more realistic based on boundary and finite elements. The forward and inverse problems in electroencephalography are discussed, including the non-uniqueness of the inverse problem. The approaches to the solution of the inverse problem described include single and multiple time-slice localization, equivalent dipole localization and the weighted minimum norm. The multiple time-slice localization approach is highlighted as probably the best available at this time and is discussed in terms of the spatiotemporal model of the EEG. The effect of noise corruption, artifacts and the number of recording electrodes on the accuracy of source localization is also mentioned. It is suggested that the main appeal of the minimum norm is that it does not assume a model for the sources and provides an estimate of the current density everywhere in the three dimensional volume of the head.

Spatial filtering of multichannel electroencephalographic recordings through principal component analysis by singular value decomposition

Principal component analysis (PCA) by singular value decomposition (SVD) may be used to analyze an epoch of a multichannel electroencephalogram (EEG) into multiple linearly independent (temporally and spatially noncorrelated) components, or features; the original epoch of the EEG may be reconstructed as a linear combination of the components. The result of SVD includes the components, expressible as time series waveforms, and the factors that determine how much each component waveform contributes to each EEG channel. By omission of some component waveforms from the linear combination, a new EEG can be reconstructed, differing from the original in useful ways. For example, artifacts can be removed and features such as ictal or interictal discharges can be enhanced by suppressing the remainder of the EEG. We developed a variation of this technique in which the factors that reconstruct the modified EEG from the original are stored as a matrix. This matrix is applied to multichannel EEG at successive times to create a new EEG continuously in real time, without redoing the time-consuming SVD. This matrix acts as a spatial filter with useful properties. We successfully applied this method to remove artifacts, including ocular movement and electrocardiographic artifacts. Removal of myogenic artifacts was much less complete, but there was significant improvement in the ability to visualize underlying activity in the presence of myogenic artifacts. The major limitations of the method are its inability to completely separate some artifacts from cerebral activity, especially when both have similar amplitudes, and the possibility that a spatial filter may distort the distribution of activities that overlap with the artifacts being removed.

Complexity of focal spikes suggests relative epileptogenicity

The EEGs of 39 children with focal or multifocal spikes were subjected to singular value decomposition (SVD) as provided by a commercial software program. We noted that in children with spikes but no clinical seizures the variance accounted for by the first component averaged 91.9%, whereas in children with seizures it was 68.0% (p < .001). The first component accounted for 85.4% in children with single spike foci, for 71.5% in those with multifocal spikes, and for 61.4% (p < 0.002) in those with both focal spikes and generalized spike-wave complexes. Spikes in the frontal and frontopolar areas were the most complex, suggesting that at least in children they tend to be the partial expression of a generalized seizure tendency rather than a result of strictly local pathology.

Spatio-temporal decomposition of the EEG: a general approach to the isolation and localization of sources

The principal-component method of source localization for the background EEG is generalized to arbitrary spatio-temporal decompositions. It is shown that as long as the spatial patterns of the decomposition span the same signal space as the principal spatial components, the computational process of attempting to localize the sources is the same. Decompositions other than the principal components are shown to be superior for the EEG in that they appear to enable individual sources to be better isolated. An example is given using the common spatial pattern decomposition and using a raw varimax rotation of a subset of the common spatial patterns. The results show that the principal component decomposition is almost ineffective for isolating spike and sharp wave activity in an EEG from a patient with epilepsy, that the common spatial pattern decomposition is significantly better and that the varimax rotation is better yet. That the varimax rotation is best is demonstrated by attempting to locate dipole sources inside the brain which account for the spike and sharp wave activity on the scalp. The question which remains is whether there exists some oblique rotation of the basis vectors of the EEG signal space which is optimal for isolating individual sources.

Principal-component localization of the sources of the background EEG

A method, based on principal components for localizing the sources of the background EEG, is presented which overcomes the previous limitations of this approach. The spatiotemporal source model of the EEG is assumed to apply, and the method involves attempting to fit the spatial aspects of this general model with an optimal rotation of a subset of the principal components of a particular EEG. The method is shown to be equivalent to the subspace scanning method, a special case of the MUSIC algorithm, which enables multiple sources to be localized individually rather than all at once. The novel aspect of the new method is that it offers a way of selecting the relevant principal components for the localization problem. The relevant principal components are chosen by decomposing the EEG using spatial patterns common with a control EEG. These spatial patterns have the property that they account for maximally different proportions of the combined variances in the two EEG's. An example is given using a particular EEG from a neurologic patient. Components containing spike and sharp wave potentials are extracted, with respect to a standard EEG derived from 15 normal volunteers. Spike and sharp wave potentials are identified visually using the common spatial patterns decomposition and an EEG reconstructed from these components. Four dipole sources are fitted to the principal components of the reconstructed EEG and these source account for over 88% of the temporal variance present in that EEG.

Spatial patterns in the background EEG underlying mental disease in man

The spatial patterns underlying differences in the background EEGs of schizophrenic, manic and depressed patients and a group of normal controls has been examined during the eyes open and eyes closed resting conditions and during 3 cognitive tasks. The method of principal-component analysis was used to extract spatial patterns which are common to the EEGs of 2 groups but which account for maximally different proportions of the combined variances. The common spatial patterns in all possible pairings of the groups were used to extract variance-related feature vectors from the individual EEG epochs in the 2 groups and the means of these vectors were subjected to statistical analyses. The results of these analyses indicate that there are significant differences in the EEGs from all 4 of the groups. The spatial patterns underlying the features which are significantly different in each comparison are shown graphically and used to suggest which brain regions might be implicated in each of the psychiatric conditions and how these are affected by the cognitive condition. The main results are that the EEGs in the schizophrenic group can be characterized by left-sided hyperactivity, in the depressed group by right-sided hyperactivity and in the manic group by bilateral hyperactivity and that these characteristics are best elicited by different cognitive states.

Systematic comparisons of interpolation techniques in topographic brain mapping

The performance of one local interpolation technique, the nearest neighbors, and two global spline techniques, one planar and the other spherical, commonly used for topographic mapping of brain potential data has been quantitatively evaluated. The method of evaluation was one of cross-validation where the potential at each site in a 31-electrode full scalp recording montage is predicted by interpolation from the other sites. Errors between the measured potentials and those predicted by interpolation were quantified using 4 measures defined as inaccuracy, precision, bias and tolerance. The evaluation was applied to the background EEGs from 5 normal volunteers and from 4 patients with epilepsy, tumor or stroke. The results indicate that none of the interpolation techniques performed well and that for localized components in the EEG, the errors can increase almost without limit. Further, the global techniques performed significantly better than the local technique with 2 being the best order for the nearest-neighbor technique and 3 for the spline techniques. It is concluded that interpolation should not be used with electrode densities of the order of that provided by the international 10-20 system neither to increase the spatial resolution of the electroencephalogram nor in more sophisticated analysis techniques in quantitative EEG for estimates such as the radial-current density.