Review on solving the inverse problem in EEG source analysis

Brain-computer interface based on generation of visual images

This paper examines the task of recognizing EEG patterns that correspond to performing three mental tasks: relaxation and imagining of two types of pictures: faces and houses. The experiments were performed using two EEG headsets: BrainProducts ActiCap and Emotiv EPOC. The Emotiv headset becomes widely used in consumer BCI application allowing for conducting large-scale EEG experiments in the future. Since classification accuracy significantly exceeded the level of random classification during the first three days of the experiment with EPOC headset, a control experiment was performed on the fourth day using ActiCap. The control experiment has shown that utilization of high-quality research equipment can enhance classification accuracy (up to 68% in some subjects) and that the accuracy is independent of the presence of EEG artifacts related to blinking and eye movement. This study also shows that computationally-inexpensive bayesian classifier based on covariance matrix analysis yields similar classification accuracy in this problem as a more sophisticated Multi-class Common Spatial Patterns (MCSP) classifier.

The adaptive pattern of the auditory N1 peak revealed by standardized low-resolution brain electromagnetic tomography

The N1 peak in the late auditory evoked potential (LAEP) decreases in amplitude following stimulus repetition, displaying an adaptive pattern. The present study explored the functional neural substrates that may underlie the N1 adaptive pattern using standardized Low Resolution Electromagnetic Tomography (sLORETA). Fourteen young normal hearing (NH) listeners participated in the study. Tone bursts (80 dB SPL) were binaurally presented via insert earphones in trains of 10; the inter-stimulus interval was 0.7s and the inter-train interval was 15s. Current source density analysis was performed for the N1 evoked by the 1st, 2nd and 10th stimuli (S(1), S(2) and S(10)) at 3 different timeframes that corresponded to the latency ranges of the N1 waveform subcomponents (70-100, 100-130 and 130-160 ms). The data showed that S(1) activated broad regions in different cortical lobes and the activation was much smaller for S(2) and S(10). Response differences in the LAEP waveform and sLORETA were observed between S(1) and S(2), but not between the S(2) and S(10). The sLORETA comparison map between S(1) and S(2) responses showed that the activation was located in the parietal lobe for the 70-100 ms timeframe, the frontal and limbic lobes for the 100-130 ms timeframe, and the frontal lobe for the 130-160 ms timeframe. These sLORETA comparison results suggest a parieto-frontal network that might help to sensitize the brain to novel stimuli by filtering out repetitive and irrelevant stimuli. This study demonstrates that sLORETA may be useful for identifying generators of scalp-recorded event related potentials and for examining the physiological features of these generators. This technique could be especially useful for cortical source localization in individuals who cannot be examined with functional magnetic resonance imaging or magnetoencephalography (e.g., cochlear implant users).

The effect of conditional probability of chord progression on brain response: an MEG study

Background

Recent electrophysiological and neuroimaging studies have explored how and where musical syntax in Western music is processed in the human brain. An inappropriate chord progression elicits an event-related potential (ERP) component called an early right anterior negativity (ERAN) or simply an early anterior negativity (EAN) in an early stage of processing the musical syntax. Though the possible underlying mechanism of the EAN is assumed to be probabilistic learning, the effect of the probability of chord progressions on the EAN response has not been previously explored explicitly.

Methodology/Principal Findings

In the present study, the empirical conditional probabilities in a Western music corpus were employed as an approximation of the frequencies in previous exposure of participants. Three types of chord progression were presented to musicians and non-musicians in order to examine the correlation between the probability of chord progression and the neuromagnetic response using magnetoencephalography (MEG). Chord progressions were found to elicit early responses in a negatively correlating fashion with the conditional probability. Observed EANm (as a magnetic counterpart of the EAN component) responses were consistent with the previously reported EAN responses in terms of latency and location. The effect of conditional probability interacted with the effect of musical training. In addition, the neural response also correlated with the behavioral measures in the non-musicians.

Conclusions/Significance

Our study is the first to reveal the correlation between the probability of chord progression and the corresponding neuromagnetic response. The current results suggest that the physiological response is a reflection of the probabilistic representations of the musical syntax. Moreover, the results indicate that the probabilistic representation is related to the musical training as well as the sensitivity of an individual.

Investigation of EEG and MEG source imaging accuracy in reconstructing extended cortical sources

Electroencephalography (EEG) and magneto-encephalography (MEG) are both currently used to reconstruct brain activity. The performance of inverse source reconstructions is dependent on the modality of signals in use as well as inverse techniques. Here we used a recently proposed sparse source imaging technique, i.e., the variation-based sparse cortical current density (VB-SCCD) algorithm to compare the use of EEG or MEG data in reconstructing extended cortical sources. We conducted Monte Carlo simulations as comparison to two other widely used source imaging techniques. The VB-SCCD technique was further evaluated in experimental EEG and MEG data. Our present results indicate that EEG and MEG have similar reconstruction performance as indicated by a metric, the area under the receiver operating characteristic curve (AUC). Furthermore, EEG and MEG have different advantages and limitations in revealing different aspects of features of extended cortical sources, which are complimentary to each other. A simultaneous EEG and MEG analysis framework is thus promising to produce much improved source reconstructions.

Amygdalar hyperactivity, a fear-related link between panic disorder and mesiotemporal epilepsy

The sudden onset, short duration and stereotyped features of panic attacks, and the fear aura of seizures starting at the mesial aspects of the temporal lobe, suggest common mechanisms underlying panic disorder (PD) and mesiotemporal epilepsy (MTLE). However, current consensus emphasizes the importance of differentiating the two entities based on 1) intact consciousness in panic attacks, 2) poor response to antiepileptics, and 3) unsuccessful electrophysiological attempts to demonstrate a relationship. We report two cases with a diagnosis of PD that had been partially responsive to first line treatments. During the EEG session, both patients developed panic symptoms with minimal EEG changes in response to paper bag-hyperventilation (PB-HV), and several minutes later presented a clear ictal EEG pattern associated with very different clinical symptoms, but both with strong fear content. Z-scored LORETA analysis showed increased current source densities (CSD) at the right amygdala in both subjects during the induced panic symptoms. Several areas were involved during the seizure, different in each subject. Yet, a very significant increase at the amygdala was found in both cases. The LORETA Z-scored source correlation (LSC) analysis also showed similar abnormal patterns during the panic symptoms in both patients, and marked differences during the seizure. These findings show a major role of amygdalar hyperactivity in both fear-related conditions for the two patients, and are discussed in relation to existing models of PD in general. Abnormal overactivation at mesiotemporal regions is poorly represented at the surface recordings but can be detected by the appropriate analytical techniques.

The role of the cingulate cortex as neural generator of the N200 and P300 in a tactile response inhibition task

Both the N200 and P300, which are, for example, evoked by Go/Nogo or Stop-Signal tasks, have long been interpreted as indicators for inhibition processes. Such interpretations have recently been challenged, and interest in the exact neural generators of these brain responses is continuously growing. Using recent methodological advancements, source estimations for the N200 and P300 as evoked by a tactile response inhibition task were computed. Current density reconstructions were also calculated accounting for interindividual differences in head geometry by incorporating information from T1-weighted magnetic resonance images. To ease comparability with relevant paradigms, the task was designed to mimic important characteristics of both Go/Nogo and Stop-Signal tasks as prototypes for a larger set of paradigms probing response inhibition. A network of neural generators was revealed, which has previously been shown to act in concert with executive control processes and thus is in full agreement with observations from other modalities. Importantly, a spatial segregation of midcingulate sources was observed. Our experimental data indicate that a left anterior region of the midcingulate cortex (MCC) is a major neural generator of the N200, whereas the midcingulate generator of the P300 is located in the right posterior MCC. Analyses of the P300 also revealed several areas, which have previously been associated with motor functions, for example, the precentral region. Our data clearly suggest a neuroanatomical and therefore also functional dissociation of the N200 and P300, a finding that cannot easily be provided by other imaging techniques.

Abnormal brain processing in hepatic encephalopathy: evidence of cerebral reorganization?

BACKGROUND AND AIM

Hepatic encephalopathy (HE) is a severe and frequent complication of liver cirrhosis characterized by abnormal cerebral function. Little is known about the underlying neural mechanisms in HE and human data are sparse. Electrophysiological methods such as evoked brain potentials after somatic stimuli can be combined with inverse modeling of the underlying brain activity. Thereby, information on neuronal dynamics and brain activity can be studied in vivo. The aim of this study was to investigate the sensory brain processing in patients with HE.

PATIENTS AND METHODS

Twelve patients with minimal or overt HE and 26 healthy volunteers were included in the study. Cerebral sensory processing was investigated as (i) an auditory reaction time task; (ii) visual and somatosensory evoked brain potentials, and (iii) reconstruction of the underlying brain activity.

RESULTS

Somatosensory evoked potentials were reproducible (all P>0.05), whereas flash evoked potentials were not reproducible (all P<0.05). Compared with healthy volunteers, the patient group had a prolonged reaction time index (P=0.03) along with increasing prolongation of latencies of median nerve evoked potentials (P<0.03). Reconstruction of the underlying brain sources showed a lateral shift in source localization of the P45 (P<0.001) and N60 components (P=0.02). A correlation between the psychometric hepatic encephalopathy score and the dipole shift corresponding to the N60 (P=0.003) component was seen.

CONCLUSION

HE patients have evidence of prolonged intracerebral nerve conduction, along with lateralization of brain activity following median nerve stimulation. This possibly represents cortical reorganization and may be important in our understanding of this condition.

Gastrointestinal symptoms in type-1 diabetes: is it all about brain plasticity?

BACKGROUND AND AIMS

Autonomic neuropathy seems to play a central role in the development of gastrointestinal symptoms in diabetes. In order to explore the neuronal mechanisms behind the symptoms we evaluated the brain processing of painful visceral stimuli.

METHODS

Evoked brain potentials were recorded to assess the response to painful oesophageal electrical stimuli in 15 healthy volunteers and 14 type-1 diabetes patients with autonomic neuropathy and related gastrointestinal symptoms. Source reconstruction analysis (fixed Multiple Signal Classification (MUSIC) algorithm) was applied to estimate the location of the evoked electrical activity in the brain.

RESULTS

The patients had increased oesophageal sensory thresholds compared to the controls (P=0.004). The latencies of the evoked brain potentials at vertex (Cz) were increased (P=0.007) and amplitudes reduced (P=0.011) in diabetics. Compared with controls the patients had a posterior shift of the electrical sources in the anterior cingulate cortex at 54 ms, and additional sources close to the posterior insula at 95 ms and in medial frontal gyrus at 184 ms.

CONCLUSIONS

There is evidence of altered central processing to visceral stimulation, and both peripheral and central mechanisms seem involved. Central neuronal reorganisation may contribute to our understanding of the gastrointestinal symptoms in patients with diabetic autonomic neuropathy and this may guide development and evaluation of new treatment modalities.

Cortical potential imaging using L-curve and GCV method to choose the regularisation parameter

BACKGROUND

The electroencephalography (EEG) is an attractive and a simple technique to measure the brain activity. It is attractive due its excellent temporal resolution and simple due to its non-invasiveness and sensor design. However, the spatial resolution of EEG is reduced due to the low conducting skull. In this paper, we compute the potential distribution over the closed surface covering the brain (cortex) from the EEG scalp potential. We compare two methods - L-curve and generalised cross validation (GCV) used to obtain the regularisation parameter and also investigate the feasibility in applying such techniques to N170 component of the visually evoked potential (VEP) data.

METHODS

Using the image data set of the visible human man (VHM), a finite difference method (FDM) model of the head was constructed. The EEG dataset (256-channel) used was the N170 component of the VEP. A forward transfer matrix relating the cortical potential to the scalp potential was obtained. Using Tikhonov regularisation, the potential distribution over the cortex was obtained.

RESULTS

The cortical potential distribution for three subjects was solved using both L-curve and GCV method. A total of 18 cortical potential distributions were obtained (3 subjects with three stimuli each - fearful face, neutral face, control objects).

CONCLUSIONS

The GCV method is a more robust method compared to L-curve to find the optimal regularisation parameter. Cortical potential imaging is a reliable method to obtain the potential distribution over cortex for VEP data.

Intrinsic dendritic filtering gives low-pass power spectra of local field potentials

The local field potential (LFP) is among the most important experimental measures when probing neural population activity, but a proper understanding of the link between the underlying neural activity and the LFP signal is still missing. Here we investigate this link by mathematical modeling of contributions to the LFP from a single layer-5 pyramidal neuron and a single layer-4 stellate neuron receiving synaptic input. An intrinsic dendritic low-pass filtering effect of the LFP signal, previously demonstrated for extracellular signatures of action potentials, is seen to strongly affect the LFP power spectra, even for frequencies as low as 10 Hz for the example pyramidal neuron. Further, the LFP signal is found to depend sensitively on both the recording position and the position of the synaptic input: the LFP power spectra recorded close to the active synapse are typically found to be less low-pass filtered than spectra recorded further away. Some recording positions display striking band-pass characteristics of the LFP. The frequency dependence of the properties of the current dipole moment set up by the synaptic input current is found to qualitatively account for several salient features of the observed LFP. Two approximate schemes for calculating the LFP, the dipole approximation and the two-monopole approximation, are tested and found to be potentially useful for translating results from large-scale neural network models into predictions for results from electroencephalographic (EEG) or electrocorticographic (ECoG) recordings.

Time and frequency domain event-related electrical activity associated with response control in schizophrenia

OBJECTIVE

To confirm previously reported abnormalities in time domain EEG components during a go/no-go task in schizophrenia, and to test the hypothesis that patients exhibit abnormalities in frequency domain components reflecting indices of behavioural impairment.

METHODS

EEG data were recorded from 17 male schizophrenia patients in a stable phase of illness and 17 healthy controls.

RESULTS

As compared with controls, patients displayed smaller N200 amplitudes and less evoked theta for correct hit trials; and smaller N200 and P300 amplitudes and less evoked delta and theta for correct reject trials. Effect sizes were largest for evoked delta. Source localisation revealed reduced activation in schizophrenia patients during the N200 and P300 time windows in anterior and posterior cingulate, medial frontal gyrus and precuneus. Evoked delta and theta oscillations were significantly correlated with the variability of reaction times and the performance level statistic d-prime.

CONCLUSIONS

The results demonstrate impairment of frontal and parietal brain areas involved in response control in schizophrenia. They also suggest that the timing of oscillations in patients is less precise leading to smaller evoked amplitudes and more variable reaction times.

SIGNIFICANCE

These findings add to the evidence that abnormal EEG oscillations contribute to impaired behavioural control in schizophrenia.

Dynamic solution to the EEG source localization problem using Kalman filters and particle filters

In this paper, we propose a solution to the EEG source localization problem considering its dynamic behavior. We assume a dipolar approach which makes the problem nonlinear. From the dynamic probabilistic model of the problem, we formulate the Extended Kalman Filter and Particle Filter solutions. In order to test the algorithms, we designed an experimental protocol based on error-related potentials. During the experiments, our dynamic solutions have allowed the estimation of sources which are varying in position and moment within the brain volume. Results confirm the activation of the anterior cingulate cortex which is the brain structure associated with error processing. These findings demonstrate the good performance of the dynamic solutions for estimating and tracking EEG neural generators.

Comparison between realistic and spherical approaches in EEG forward modelling

In electroencephalography (EEG) a valid conductor model of the head (forward model) is necessary for predicting measurable scalp voltages from intra-cranial current distributions. All inverse models, capable of inferring the spatial distribution of the neural sources generating measurable electrical and magnetic signals outside the brain are normally formulated in terms of a pre-estimated forward model, which implies considering one (or more) current dipole(s) inside the head and computing the electrical potentials generated at the electrode sites on the scalp surface. Therefore, the accuracy of the forward model strongly affects the reliability of the source reconstruction process independently of the specific inverse model. So far, it is as yet unclear which brain regions are more sensitive to the choice of different model geometry, from both quantitative and qualitative points of view. In this paper, we compare the finite difference method-based realistic model with the four-layers sensor-fitted spherical model using simulated cortical sources in the MNI152 standard space. We focused on the investigation of the spatial variation of the lead fields produced by simulated cortical sources which were placed on the reconstructed mesh of the neocortex along the surface electrodes of a 62-channel configuration. This comparison is carried out by evaluating a point spread function all over the brain cortex, with the aim of finding the lead fields mismatch between realistic and spherical geometry. Realistic geometry turns out to be a relevant factor of improvement which is particularly important when considering sources placed in the temporal or in the occipital cortex. In these situations, using a realistic head model will allow a better spatial discrimination of neural sources when compared to the spherical model.

Estimation of dynamic neural activity using a Kalman filter approach based on physiological models

This paper presents a new method to estimate dynamic neural activity from EEG signals. The method is based on a Kalman filter approach, using physiological models that take both spatial and temporal dynamics into account. The filter's performance (in terms of estimation error) is analyzed for the cases of linear and nonlinear models having either time invariant or time varying parameters. The best performance is achieved with a nonlinear model with time-varying parameters.

Improving human brain mapping via joint inversion of brain electrodynamics and the BOLD signal

We present several methods to improve the resolution of human brain mapping by combining information obtained from surface electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) of the same participants performing the same task in separate imaging sessions. As an initial step in our methods we used independent component analysis (ICA) to obtain task-related sources for both EEG and fMRI. We then used that information in an integrated cost function that attempts to match both data sources and trades goodness of fit in one regime for another. We compared the performance and drawbacks of each method in localizing sources for a dual visual evoked response experiment, and we contrasted the results of adding fMRI information to simple EEG-only inversion methods. We found that adding fMRI information in a variety of ways gives superior results to classical minimum norm source estimation. Our findings lead us to favor a method which attempts to match EEG scalp dynamics along with voxel power obtained from ICA-processed blood oxygenation level dependent (BOLD) data; this method of joint inversion enables us to treat the two data sources as symmetrically as possible.