A probabilistic algorithm integrating source localization and noise suppression for MEG and EEG data

We have developed a novel probabilistic model that estimates neural source activity measured by MEG and EEG data while suppressing the effect of interference and noise sources. The model estimates contributions to sensor data from evoked sources, interference sources and sensor noise using Bayesian methods and by exploiting knowledge about their timing and spatial covariance properties. Full posterior distributions are computed rather than just the MAP estimates. In simulation, the algorithm can accurately localize and estimate the time courses of several simultaneously active dipoles, with rotating or fixed orientation, at noise levels typical for averaged MEG data. The algorithm even performs reasonably at noise levels typical of an average of just a few trials. The algorithm is superior to beamforming techniques, which we show to be an approximation to our graphical model, in estimation of temporally correlated sources. Success of this algorithm using MEG data for localizing bilateral auditory cortex, low-SNR somatosensory activations, and for localizing an epileptic spike source are also demonstrated.     

Probabilistic algorithms for MEG/EEG source reconstruction using temporal basis functions learned from data

We present two related probabilistic methods for neural source reconstruction from MEG/EEG data that reduce effects of interference, noise, and correlated sources. Both methods localize source activity using a linear mixture of temporal basis functions (TBFs) learned from the data. In contrast to existing methods that use predetermined TBFs, we compute TBFs from data using a graphical factor analysis based model [Nagarajan, S.S., Attias, H.T., Hild, K.E., Sekihara, K., 2007a. A probabilistic algorithm for robust interference suppression in bioelectromagnetic sensor data. Stat Med 26, 3886-3910], which separates evoked or event-related source activity from ongoing spontaneous background brain activity. Both algorithms compute an optimal weighting of these TBFs at each voxel to provide a spatiotemporal map of activity across the brain and a source image map from the likelihood of a dipole source at each voxel. We explicitly model, with two different robust parameterizations, the contribution from signals outside a voxel of interest. The two models differ in a trade-off of computational speed versus accuracy of learning the unknown interference contributions. Performance in simulations and real data, both with large noise and interference and/or correlated sources, demonstrates significant improvement over existing source localization methods.     

Cortical source estimates of gamma band amplitude and phase are different in schizophrenia

Reductions in gamma band phase synchrony and evoked power have been reported in schizophrenic subjects in response to auditory stimuli. These results have been observed in the EEG at one or two electrode sites. We wished to extend these results using magnetic field data to estimate the responses at the neural generators themselves in each hemisphere. Whole head magnetoencephalographic (MEG) recordings were used to estimate the phase and amplitude behavior of sources in primary auditory cortex in both hemispheres of schizophrenic and comparison subjects. Both ipsi- and contralateral cases were evaluated using a driving (40 Hz modulated 1 kHz carrier) and a non-driving (1 kHz tone) stimulus. We used source space projection (SSP) to collapse the magnetic field data into estimates of the time course of source strengths in individual trials. Complex wavelet based time–frequency decomposition was used to compute inter-trial phase locking factor (PLF), and mean evoked and induced amplitude for each cortical generator. Schizophrenic subjects showed reduced SSP PLF and evoked source strength for contralateral generators responding to the driving stimulus in both hemispheres. For the pure tone stimulus, only the left hemisphere PLF's in the transient window were reduced. In contrast, subjects with schizophrenia exhibited higher induced 40 Hz power to both stimulus types, consistent with the reduced PLF findings. The method of SSP combined with wavelet based complex demodulation produces a significant improvement in signal-to-noise ratio, and directly estimates the activity of the cortical generators responsible for gamma band auditory MEG evoked fields. Schizophrenic subjects exhibit significant impairment of generation and phase locking of this activity in auditory cortex, suggesting an impairment of GABA-ergic inhibitory interneuronal modulation of pyramidal cell activity.     

Spectral spatiotemporal imaging of cortical oscillations and interactions in the human brain

This paper presents a computationally efficient source estimation algorithm that localizes cortical oscillations and their phase relationships. The proposed method employs wavelet-transformed magnetoencephalography (MEG) data and uses anatomical MRI to constrain the current locations to the cortical mantle. In addition, the locations of the sources can be further confined with the help of functional MRI (fMRI) data. As a result, we obtain spatiotemporal maps of spectral power and phase relationships. As an example, we show how the phase locking value (PLV), that is, the trial-by-trial phase relationship between the stimulus and response, can be imaged on the cortex. We apply the method to spontaneous, evoked, and driven cortical oscillations measured with MEG. We test the method of combining MEG, structural MRI, and fMRI using simulated cortical oscillations along Heschl's gyrus (HG). We also analyze sustained auditory gamma-band neuromagnetic fields from MEG and fMRI measurements. Our results show that combining the MEG recording with fMRI improves source localization for the non-noise-normalized wavelet power. In contrast, noise-normalized spectral power or PLV localization may not benefit from the fMRI constraint. We show that if the thresholds are not properly chosen, noise-normalized spectral power or PLV estimates may contain false (phantom) sources, independent of the inclusion of the fMRI prior information. The proposed algorithm can be used for evoked MEG/EEG and block-designed or event-related fMRI paradigms, or for spontaneous MEG data sets. Spectral spatiotemporal imaging of cortical oscillations and interactions in the human brain can provide further understanding of large-scale neural activity and communication between different brain regions.     

Spatiotemporal dynamics of bilingual word processing

Studies with monolingual adults have identified successive stages occurring in different brain regions for processing single written words. We combined magnetoencephalography and magnetic resonance imaging to compare these stages between the first (L1) and second (L2) languages in bilingual adults. L1 words in a size judgment task evoked a typical left-lateralized sequence of activity first in ventral occipitotemporal cortex (VOT: previously associated with visual word-form encoding) and then ventral frontotemporal regions (associated with lexico-semantic processing). Compared to L1, words in L2 activated right VOT more strongly from ～ 135 ms; this activation was attenuated when words became highly familiar with repetition. At ～ 400 ms, L2 responses were generally later than L1, more bilateral, and included the same lateral occipitotemporal areas as were activated by pictures. We propose that acquiring a language involves the recruitment of right hemisphere and posterior visual areas that are not necessary once fluency is achieved.     

Localization of human supratemporal auditory areas from intracerebral auditory evoked potentials using distributed source models

While source localization methods are increasingly developed to identify brain areas underlying scalp electro/magnetoencephalographic data (EEG/MEG), these methods have not yet been used to identify the sources of intracerebral signals which offer highly detailed information. Here, we adapted the minimum current estimates method to intracranial data in order to localize supratemporal sources of intracerebral auditory 1-kHz-tone-evoked potentials occurring within 100 ms after stimulus onset. After an evaluation of localization method and despite inter-subject variability, we found a common spatiotemporal pattern of activities, which involved the first Heschl's gyrus (H1) and sulcus (HS), the Planum Temporale (PT), H2/H3 when present, and the superior temporal gyrus (STG). Four time periods of activity were distinguished, corresponding to the time range of the scalp components P0, Na, Pa/Pb, and N100. The sources of the earliest components P0 (16-19 ms) and Na (20-25 ms) could be identified in the postero-medial portion of HS or H1. Then, several areas became simultaneously active after 25 ms. The Pa/Pb time range (30-50 ms) was characterized by a medio-lateral and postero-anterior propagation of activity over the supratemporal plane involving successively H1/HS, the Planum Temporale, H2/H3 when present, and the STG. Finally, we found to a large extent that the N100 (55-100 ms) involved almost the same areas as those active during the Pa/Pb complex, with a similar propagation of activities. Reconstructing scalp data from these sources on fictive EEG/MEG channels reproduced classical auditory evoked waveforms and topographies. In conclusion, the spatiotemporal pattern of activation of supratemporal auditory areas could be identified on the individual anatomy using current estimates from intracerebral data. Such detailed localization approach could also be used prior to epilepsy surgery to help identify epileptogenic foci and preserve functional cortical areas.     

Differential brain activation associated with laser-evoked burning and pricking pain: An event-related fMRI study

An important question remains as to how the brain differentially processes first (pricking) pain mediated by Aδ-nociceptors versus second (burning) pain mediated by C-nociceptors. In the present cross-over randomized, within-subjects controlled study, brain activity patterns were examined with event-related fMRI while pricking and burning pain were selectively evoked using a diode laser. Stimuli evoking equivalent pain intensities were delivered to the dorsum of the left foot. Different laser parameters were used to elicit pricking (60 ms pulse duration) and burning (2.0 s pulse duration) pain. Whole brain group analysis showed that several brain areas were commonly activated by pricking and burning pain, including bilateral thalamus, bilateral anterior insula, bilateral posterior parietal lobule, contralateral dorsolateral prefrontal cortex, ipsilateral cerebellum, and mid anterior cingulate cortex. These findings show that pricking and burning pain were associated with activity in many of the same nociceptive processing brain regions. This may be expected given that Aδ-and C-nociceptive signals converge to a great extent at the level of the dorsal horn. Other brain regions showed differential processing. Stronger activation in the pricking pain condition was found in the ipsilateral hippocampus, bilateral parahippocampal gyrus, bilateral fusiform gyrus, contralateral cerebellum and contralateral cuneus/parieto-occipital sulcus. Stronger activation in the burning pain condition was found in the ipsilateral dorsolateral prefrontal cortex. These differential activation patterns suggest preferential importance of Aδ-fiber signals versus C-fiber signals for these specific brain regions.     

A graphical model for estimating stimulus-evoked brain responses from magnetoencephalography data with large background brain activity

This paper formulates a novel probabilistic graphical model for noisy stimulus-evoked MEG and EEG sensor data obtained in the presence of large background brain activity. The model describes the observed data in terms of unobserved evoked and background factors with additive sensor noise. We present an expectation maximization (EM) algorithm that estimates the model parameters from data. Using the model, the algorithm cleans the stimulus-evoked data by removing interference from background factors and noise artifacts and separates those data into contributions from independent factors. We demonstrate on real and simulated data that the algorithm outperforms benchmark methods for denoising and separation. We also show that the algorithm improves the performance of localization with beamforming algorithms.     

Two electrophysiological stages of spatial orienting towards fearful faces: early temporo-parietal activation preceding gain control in extrastriate visual cortex

Visuo-spatial attention tends to be prioritized towards emotionally negative stimuli such as fearful faces, as opposed to neutral or positive stimuli. Using a covert orienting task, we previously showed that a lateral occipital P1 component, with extrastriate neural sources, was selectively enhanced to lateralized visual targets replacing a fearful face (fear-valid trial) than the same targets replacing a neutral face (fear-invalid trial), providing evidence for exogenous spatial orienting of attention towards threat cues. Here, we describe a new analysis of these data, using topographic evoked potentials mapping methods combined with a distributed source localization technique. We show that an early field topography (40-80 ms post-target onset) with a centro-parietal negativity and a left posterior parietal source distinguished fear-valid from fear-invalid trials, whereas a distinct activity with anterior cingulate sources was selectively evoked during fear-invalid trials. At the same latency, or later, no difference in field topography was found for valid compared to invalid trials with happy faces. The early parietal map preceded a modulation in amplitude of the field strength (approximately 130 ms), corresponding to the enhanced lateral occipital P1 during valid trials in the fear condition. Furthermore, this early topography at 40-80 ms was positively correlated with the subsequent amplitude modulation of P1 at 130-160 ms in the fear condition, suggesting a possible functional coupling between these two successive events. These data have important implications for models of spatial attention and interactions with emotion. They suggest two successive stages of neural activity during exogenous orienting of attention towards visual targets following fearful faces, including an early posterior parietal negativity, followed by gain control mechanisms enhancing visual responses in extrastriate occipital cortex.     

Robust Bayesian estimation of the location, orientation, and time course of multiple correlated neural sources using MEG

The synchronous brain activity measured via MEG (or EEG) can be interpreted as arising from a collection (possibly large) of current dipoles or sources located throughout the cortex. Estimating the number, location, and time course of these sources remains a challenging task, one that is significantly compounded by the effects of source correlations and unknown orientations and by the presence of interference from spontaneous brain activity, sensor noise, and other artifacts. This paper derives an empirical Bayesian method for addressing each of these issues in a principled fashion. The resulting algorithm guarantees descent of a cost function uniquely designed to handle unknown orientations and arbitrary correlations. Robust interference suppression is also easily incorporated. In a restricted setting, the proposed method is shown to produce theoretically zero reconstruction error estimating multiple dipoles even in the presence of strong correlations and unknown orientations, unlike a variety of existing Bayesian localization methods or common signal processing techniques such as beamforming and sLORETA. Empirical results on both simulated and real data sets verify the efficacy of this approach.     

Head movements of children in MEG: quantification, effects on source estimation, and compensation

Head movements during magnetoencephalography (MEG) recordings may lead to inaccurate localization of brain activity. This can be particularly problematic for studies with children. We quantified head movements in 8- to 12-year-old children performing a cognitive task and examined how the movements affected source estimation. Each child was presented auditory word stimuli in five 4-min runs. The mean change in the MEG sensor locations during the experiment ranged from 3 to 26mm across subjects. The variation in the head position was largest in the up-down direction. The mean localization error in equivalent current dipole (ECD) simulations was 12mm for runs with the most head movement, with the frontal cortex appearing to be most prone to errors due to head movements. In addition, we examined the effect of head movements on two types of source estimates, ECDs and minimum-norm estimates (MNE), for an auditory evoked response. Application of a recently introduced signal space separation (SSS) method to compensate for the head movements was found to increase the goodness-of-fit of the ECDs, reduce the spatial confidence intervals of the ECDs, and enhance the peak amplitude in the MNE. These results are indicative of the SSS method being able to compensate for the spatial smoothing of the signals caused by head movements. Overall, the results suggest that MEG source estimates are relatively robust against head movements in children, and that confounds due to head movements can be successfully dealt with in MEG studies of developmental cognition.     

Application of multi-source minimum variance beamformers for reconstruction of correlated neural activity

Linearly constrained minimum variance beamformers are highly effective for analysis of weakly correlated brain activity, but their performance degrades when correlations become significant. Multiple constrained minimum variance (MCMV) beamformers are insensitive to source correlations but require a priori information about the source locations. Besides the question whether unbiased estimates of source positions and orientations can be obtained remained unanswered. In this work, we derive MCMV-based source localizers that can be applied to both induced and evoked brain activity. They may be regarded as a generalization of scalar minimum-variance beamformers for the case of multiple correlated sources. We show that for arbitrary noise covariance these beamformers provide simultaneous unbiased estimates of multiple source positions and orientations and remain bounded at singular points. We also propose an iterative search algorithm that makes it possible to find sources approximately without a priori assumptions about their locations and orientations. Simulations and analyses of real MEG data demonstrate that presented approach is superior to traditional single-source beamformers in situations where correlations between the sources are significant.     

The Role of the Thalamus in “Top Down” Modulation of Attention to Sound

By correlating rCBF with rate of presentation of tones we used PET to identify brain regions where auditory signals elicited a transient neural response. In one condition volunteers were asked to attend to the tones and ignore visual signals, while in the second condition they were asked to attend to the visual signals and ignore the tones. Activity in primary auditory cortex and adjacent areas was strongly correlated with rate of tone presentation, but this relationship was not affected by the direction of attention. In only one area, the right midthalamus, was the response to tones modulated by attention. In this area responses to tones occurred when attention was directed to sound, but not when attention was directed to visual stimuli. There is considerable evidence that the EEG evoked response to tones (N100/Nd response) is strongly modulated by attention and arises in auditory cortex. The ERP is the sum of activity from many sources. The amplitude of this response reflects not only the amount of activity in these sources, but also the degree of synchrony between them. The difference between these typical ERP results and our result from PET could be resolved if we assume that, in our paradigm, attention did not increase the amount of neural activity in auditory cortex, but rather the degree of synchrony between many sources. The signal in the thalamus, which we observed only when volunteers were attending to the tones, might provide the basis for this synchrony.     

Noxious stimulation in children receiving general anaesthesia evokes an increase in delta frequency brain activity

More than 235,000 children/year in the UK receive general anaesthesia, but it is unknown whether nociceptive stimuli alter cortical brain activity in anaesthetised children. Time-locked electroencephalogram (EEG) responses to experimental tactile stimuli, experimental noxious stimuli, and clinically required cannulation were examined in 51 children (ages 1–12 years) under sevoflurane monoanaesthesia. Based on a pilot study (n = 12), we hypothesised that noxious stimulation in children receiving sevoflurane monoanaesthesia would evoke an increase in delta activity. This was tested in an independent sample of children (n = 39), where a subset (n = 11) had topical local anaesthetic applied prior to stimulation. A novel method of time-locking the stimuli to the EEG recording was developed using an event detection interface and high-speed camera. Clinical cannulation evoked a significant increase (34.2 ± 8.3%) in delta activity (P = 0.042), without concomitant changes in heart rate or reflex withdrawal, which was not observed when local anaesthetic was applied (P = 0.30). Experimental tactile (P = 0.012) and noxious (P = 0.0099) stimulation also evoked significant increases in delta activity, but the magnitude of the response was graded with stimulus intensity, with the greatest increase evoked by cannulation. We demonstrate that experimental and clinically essential noxious procedures, undertaken in anaesthetised children, alter the pattern of EEG activity, that this response can be inhibited by local anaesthetic, and that this measure is more sensitive than other physiological indicators of nociception. This technique provides the possibility that sensitivity to noxious stimuli during anaesthesia could be investigated in other clinical populations.     

Performance evaluation of the Champagne source reconstruction algorithm on simulated and real M/EEG data

In this paper, we present an extensive performance evaluation of a novel source localization algorithm, Champagne. It is derived in an empirical Bayesian framework that yields sparse solutions to the inverse problem. It is robust to correlated sources and learns the statistics of non-stimulus-evoked activity to suppress the effect of noise and interfering brain activity. We tested Champagne on both simulated and real M/EEG data. The source locations used for the simulated data were chosen to test the performance on challenging source configurations. In simulations, we found that Champagne outperforms the benchmark algorithms in terms of both the accuracy of the source localizations and the correct estimation of source time courses. We also demonstrate that Champagne is more robust to correlated brain activity present in real MEG data and is able to resolve many distinct and functionally relevant brain areas with real MEG and EEG data.     

Therapy-related reorganization of language in both hemispheres of patients with chronic aphasia

The brain processes of language recovery after stroke are poorly understood, partly because past research did not allow to differentiate the effects of spontaneous restitution processes from those of learning-related cortical reorganization. Here, we use a new approach offered by recently developed intense neuropsychological therapy methods, which allow for improving language functions within a short time period. Stroke patients with chronic aphasia received intense language therapy for 2 weeks and, over this period, improved their language performance as assessed using clinical tests. Neurophysiological activity elicited by words and pseudowords was measured before and after treatment. Over the therapy interval, early word evoked potentials (latency 250-300 ms) became significantly stronger whereas pseudoword responses did not change. Word-specific changes were documented by analyses of ERP amplitudes and root mean square values, which revealed interactions of the factors Assessment time (before vs. after therapy) and Wordness (word vs. pseudoword). Source localization using Minimum Norm Current Estimates showed that bilateral cortical sources activated by word stimuli contributed to the change, suggesting that neuronal networks distributed over both hemispheres are the substrate of cortical reorganization of language processing in intense aphasia therapy. Word-evoked differences in source strengths were significantly correlated with performance on a clinical language test, demonstrating a link between behavioral and neurophysiological changes. We suggest that the early word-evoked negativity might represent an index of reorganization of language after stroke and thus an aphasia recovery potential.     

Brain lesion diagnosis by recording cochlear and brainstem responses to sound stimuli

A technique has been developed for the recording of the cochlear action potential (electrocochleography) and the brainstem evoked responses to click stimuli by means of earlobe and scalp electrodes with an average response computer. This technique has already proved its usefulness in diagnosis of hearing loss. Since the brainstem responses are generated in the successive brainstem auditory nuclei and since the auditory nuclei and pathways constitute a relatively large volume of brainstem tissue, there is reason to believe that this same technique can also contribute to the diagnosis of brain stem lesions and their localization. When these recordings were made in patients with clinical signs of brain stem involvement, one (or more) of the usual response waves was smaller in amplitude, prolonged in latency, or completely absent. Recording of the cochlear and brainstem evoked potentials thus seems to be a new, simple and rapid tool for the diagnosis of brainstem lesions.     

The path to success in auditory spatial discrimination: electrical neuroimaging responses within the supratemporal plane predict performance outcome

Auditory scene analysis requires the accurate encoding and comparison of the perceived spatial positions of sound sources. The electrophysiological correlates of auditory spatial discrimination and their relationship to performance accuracy were studied in humans by applying electrical neuroimaging analyses to auditory evoked potentials (AEPs) that were recorded during the completion of a near-threshold S1-S2 paradigm within the right hemispace. Data were sorted as a function of performance accuracy, and AEP responses 75-117 ms after the presentation of the first sound differed topographically between trials leading to correct and incorrect spatial discrimination. Distributed source estimations revealed that this followed from significantly stronger activity within the left (i.e. contralateral) supratemporal plane (STP) and the left inferior parietal lobule prior to correct versus incorrect discrimination performance. Successful spatial discrimination thus depends on the activity of distinct configurations of active brain networks within the contralateral temporo-parietal cortex over a time period when the first sound position is being encoded. Furthermore, significant positive correlations were observed between performance accuracy and the intracranial activity estimated within the left STP. The efficacy of S1 processing within the STP is thus predictive of behavioral performance outcome during auditory spatial discrimination. Our data support a model wherein refinement of spatial representations occurs within the STP and that interactions with parietal structures allow for transformations into coordinate frames that are required for higher-order computations including absolute localization of sound sources.     

EEG signatures of auditory activity correlate with simultaneously recorded fMRI responses in humans

We recorded auditory-evoked potentials (AEPs) during simultaneous, continuous fMRI and identified trial-to-trial correlations between the amplitude of electrophysiological responses, characterised in the time domain and the time–frequency domain, and the hemodynamic BOLD response. Cortical AEPs were recorded from 30 EEG channels within the 3 T MRI scanner with and without the collection of simultaneous BOLD fMRI. Focussing on the Cz (vertex) EEG response, single-trial AEP responses were measured from time-domain waveforms. Furthermore, a novel method was used to characterise the single-trial AEP response within three regions of interest in the time–frequency domain (TF-ROIs). The latency and amplitude values of the N1 and P2 AEP peaks during fMRI scanning were not significantly different from the Control session (p > 0.16). BOLD fMRI responses to the auditory stimulation were observed in bilateral secondary auditory cortices as well as in the right precentral and postcentral gyri, anterior cingulate cortex (ACC) and supplementary motor cortex (SMC). Significant single-trial correlations were observed with a voxel-wise analysis, between (1) the magnitude of the EEG TF-ROI1 (70–800 ms post-stimulus, 1–5 Hz) and the BOLD response in right primary (Heschl's gyrus) and secondary (STG, planum temporale) auditory cortex; and (2) the amplitude of the P2 peak and the BOLD response in left pre- and postcentral gyri, the ACC and SMC. No correlation was observed with single-trial N1 amplitude on a voxel-wise basis. An fMRI-ROI analysis of functionally-identified auditory responsive regions identified further single-trial correlations of BOLD and EEG responses. The TF amplitudes in TF-ROI1 and TF-ROI2 (20–400 ms post-stimulus, 5–15 Hz) were significantly correlated with the BOLD response in all bilateral auditory areas investigated (planum temporale, superior temporal gyrus and Heschl's gyrus). However the N1 and P2 peak amplitudes, occurring at similar latencies did not show a correlation in these regions. N1 and P2 peak amplitude did correlate with the BOLD response in bilateral precentral and postcentral gyri and the SMC. Additionally P2 and TF-ROI1 both correlated with the ACC. TF-ROI3 (400–900 ms post-stimulus, 4–10 Hz) correlations were only observed in the ACC and SMC. Across the group, the subject-mean N1 peak amplitude correlated with the BOLD response amplitude in the primary and secondary auditory cortices bilaterally, as well as the right precentral gyrus and SMC. We confirm that auditory-evoked EEG responses can be recorded during continuous and simultaneous fMRI. We have presented further evidence of an empirical single-trial coupling between the EEG and BOLD fMRI responses, and show that a time–frequency decomposition of EEG signals can yield additional BOLD fMRI correlates, predominantly in auditory cortices, beyond those found using the evoked response amplitude alone.