Investigations of dipole localization accuracy in MEG using the bootstrap

We describe the use of the nonparametric bootstrap to investigate the accuracy of current dipole localization from magnetoencephalography (MEG) studies of event-related neural activity. The bootstrap is well suited to the analysis of event-related MEG data since the experiments are repeated tens or even hundreds of times and averaged to achieve acceptable signal-to-noise ratios (SNRs). The set of repetitions or epochs can be viewed as a set of independent realizations of the brain's response to the experiment. Bootstrap resamples can be generated by sampling with replacement from these epochs and averaging. In this study, we applied the bootstrap resampling technique to MEG data from somatotopic experimental and simulated data. Four fingers of the right and left hand of a healthy subject were electrically stimulated, and about 400 trials per stimulation were recorded and averaged in order to measure the somatotopic mapping of the fingers in the S1 area of the brain. Based on single-trial recordings for each finger we performed 5000 bootstrap resamples. We reconstructed dipoles from these resampled averages using the Recursively Applied and Projected (RAP)-MUSIC source localization algorithm. We also performed a simulation for two dipolar sources with overlapping time courses embedded in realistic background brain activity generated using the prestimulus segments of the somatotopic data. To find correspondences between multiple sources in each bootstrap, sample dipoles with similar time series and forward fields were assumed to represent the same source. These dipoles were then clustered by a Gaussian Mixture Model (GMM) clustering algorithm using their combined normalized time series and topographies as feature vectors. The mean and standard deviation of the dipole position and the dipole time series in each cluster were computed to provide estimates of the accuracy of the reconstructed source locations and time series.     

Bayesian analysis of the neuromagnetic inverse problem with l(p)-norm priors

Magnetoencephalography (MEG) allows millisecond-scale non-invasive measurement of magnetic fields generated by neural currents in the brain. However, localization of the underlying current sources is ambiguous due to the so-called inverse problem. The most widely used source localization methods (i.e., minimum-norm and minimum-current estimates (MNE and MCE) and equivalent current dipole (ECD) fitting) require ad hoc determination of the cortical current distribution (l(2)-, l(1)-norm priors and point-sized dipolar, respectively). In this article, we perform a Bayesian analysis of the MEG inverse problem with l(p)-norm priors for the current sources. This way, we circumvent the arbitrary choice between l(1)- and l(2)-norm prior, which is instead rendered automatically based on the data. By obtaining numerical samples from the joint posterior probability distribution of the source current parameters and model hyperparameters (such as the l(p)-norm order p) using Markov chain Monte Carlo (MCMC) methods, we calculated the spatial inverse estimates as expectation values of the source current parameters integrated over the hyperparameters. Real MEG data and simulated (known) source currents with realistic MRI-based cortical geometry and 306-channel MEG sensor array were used. While the proposed model is sensitive to source space discretization size and computationally rather heavy, it is mathematically straightforward, thus allowing incorporation of, for instance, a priori functional magnetic resonance imaging (fMRI) information.     

The use of anatomical constraints with MEG beamformers

Synthetic Aperture Magnetometry (SAM) is a beamformer approach for the localisation of neuronal activity from EEG/MEG data. SAM estimates the optimum orientation of each source in a predefined source space by a nonlinear search for the orientation that maximises the beamformer output. However, MEG is most sensitive to cortical sources and these sources are generally oriented perpendicular to the surface. The reconstructed neuronal activity can therefore reasonably be constrained to the cortical surface, orientated perpendicular to it, therefore removing the search for the optimum orientation for the computation of the beamformer weights. This paper sets out to compare the performance of a constrained and unconstrained beamformer (SAM), with respect to the localisation accuracy of the source reconstructions and the spatial resolution. Fifty sources were randomly placed on a cortical surface estimated from an MRI, and we simulated data over a range of different signal-to-noise ratios (SNRs) for each source. These datasets were analysed using both an unconstrained beamformer (SAM) and a constrained beamformer (with the sources orientated perpendicular to the cortical surface). The influence of errors in the estimation of the surface location and surface normals on the performance of the constrained beamformer, representing MEG/MRI coregistration and segmentation errors, were also examined. The spatial resolution of the beamformer improves, typically by a factor of four by applying anatomical constraints, and the localisation accuracy improves marginally. However, the advantage in spatial resolution disappears when errors are introduced into the orientation and location constraints, and, moreover, the localisation accuracy of the inaccurately constrained beamformer degrades rapidly. We conclude that the use of anatomical constraints is only advantageous if the MEG/MRI coregistration error is smaller than 2 mm and the error in the estimation of the cortical surface orientation is smaller than 10 degrees.     

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.     

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.     

Quantification of the benefit from integrating MEG and EEG data in minimum l2-norm estimation

Source current estimation from electromagnetic (MEG and EEG) signals is an ill-posed problem that often produces blurry or inaccurately positioned estimates. The two modalities have distinct factors limiting the resolution, e.g., MEG cannot detect radially oriented sources, while EEG is sensitive to accuracy of the head model. This makes combined EEG + MEG estimation techniques desirable, but different acquisition noise statistics, complexity of the head models, and lack of pertinent metrics all complicate the assessment of the resulting improvements. We investigated analytically the effect of including EEG recordings in MEG studies versus the addition of new MEG channels when computing noise-normalized minimum ℓ₂-norm estimates. Three-compartment boundary-element forward models were constructed using structural MRI scans for four subjects. Singular value analysis of the resulting forward models predicted better performance of the EEG + MEG case in the form of higher matrix rank. MNE inverse operators for EEG, MEG and EEG + MEG were constructed using the sensor noise covariance estimated from data. Metrics derived from the resolution matrices predicted higher spatial resolution in EEG + MEG as compared to MEG due to decreased spread (lower spatial dispersion, higher resolution index) with no reduction in dipole localization error. The effect was apparent in all source locations, with increased magnitude for deep areas such as the cingulate cortex. We were also able to corroborate the results for the somatosensory cortex using median nerve responses.     

A new method to identify multiple sources of oscillatory activity from magnetoencephalographic data

Identifying the sources of oscillatory activity in the human brain is a challenging problem in current magnetoencephalography (MEG) and electroencephalography (EEG) research. The fluctuations in phase and amplitude of cortical oscillations preclude signal averaging over successive sections of the data without a priori assumptions. In addition, several sources at different locations often produce oscillatory activity at similar frequencies. For example, spontaneous oscillatory activity in the 8- to 13-Hz band is produced simultaneously at least in the posterior parts of the brain and bilaterally in the sensorimotor cortices. The previous approaches of identifying sources of oscillatory activity by dipole modeling of bandpass filtered data are quite laborious and require that multiple criteria are defined by an experienced user. In this work we introduce a convenient method for source localization using minimum current estimates in the frequency domain. Individual current estimates are calculated for the Fourier transforms of successive sections of continuous data. These current estimates are then averaged. The algorithm was tested on simulated and measured MEG data and compared with conventional dipole modeling. The main advantage of the proposed method is that it provides an efficient approach for simultaneous estimation of multiple sources of oscillatory activity in the same frequency band.     

Comparison of noise-normalized minimum norm estimates for MEG analysis using multiple resolution metrics

Noise-normalization has been shown to partly compensate for the localization bias towards superficial sources in minimum norm estimation. However, it has been argued that in order to make inferences for the case of multiple sources, localization properties alone are insufficient. Instead, multiple measures of resolution should be applied to both point-spread and cross-talk functions (PSFs and CTFs). Here, we demonstrate that noise-normalization affects the shapes of PSFs, but not of CTFs. We evaluated PSFs and CTFs for the MNE, dSPM and sLORETA inverse operators, on the metrics dipole localization error (DLE), spatial dispersion (SD) and overall amplitude (OA). We used 306-channel MEG configurations obtained from 17 subjects in a real experiment, including individual noise covariance matrices and head geometries. We confirmed that for PSFs DLE improved after noise normalization, and is zero for sLORETA. However, SD was generally lower for the unnormalized MNE. OA distributions were similar for all three methods, indicating that all three methods may greatly underestimate some sources relative to others. The reliability of differences between methods across subjects was demonstrated using distributions of standard deviations and p-values from paired t-tests. As predicted, the shapes of CTFs were the same for all methods, reflecting the general resolution limits of the inverse problem. This means that noise-normalization is of no consequence where linear estimation procedures are used as "spatial filters." While low DLE is advantageous for the localization of a single source, or possibly a few spatially distinct sources, the benefit for the case of complex source distributions is not obvious. We suggest that software packages for source estimation should include comprehensive tools for evaluating the performance of different methods.     

The influence of brain tissue anisotropy on human EEG and MEG.

The influence of gray and white matter tissue anisotropy on the human electroencephalogram (EEG) and magnetoencephalogram (MEG) was examined with a high resolution finite element model of the head of an adult male subject. The conductivity tensor data for gray and white matter were estimated from magnetic resonance diffusion tensor imaging. Simulations were carried out with single dipoles or small extended sources in the cortical gray matter. The inclusion of anisotropic volume conduction in the brain was found to have a minor influence on the topology of EEG and MEG (and hence source localization). We found a major influence on the amplitude of EEG and MEG (and hence source strength estimation) due to the change in conductivity and the inclusion of anisotropy. We expect that inclusion of tissue anisotropy information will improve source estimation procedures.     

Towards source volume estimation of interictal spikes in focal epilepsy using magnetoencephalography

Interictal spikes are a hallmark of cortical epileptogenicity; their spatial distribution in the cortex defines the so-called ‘irritative’ zone or spiking volume (SV). Delineating the SV precisely is a challenge during the presurgical evaluation of patients with epilepsy. Magnetoencephalography (MEG) recordings enable determination of the brain sources of epileptic spikes using source localization procedures. Most previous clinical MEG studies have relied on dipole modeling of epileptic spikes, which does not permit a volumetric estimation of the spiking cortex.In the present study, we propose a new source modeling procedure, Volumetric Imaging of Epileptic Spikes (VIES). In VIES, the SV is identified as the 3D region where sources of the high frequency activities (> 20 Hz) associated with epileptic spikes are distributed. We localized these sources using a beamforming approach (DICS, Dynamic Imaging of Coherent Neural Sources). To determine the optimal parameters and accuracy of the method, we compared the SV obtained by VIES with the SV defined by the invasive gold standard, intracranial stereotactic EEG recordings (SEEG), in 21 patients with focal epilepsy. Using rigorous validation criteria based on the exact anatomical location of SEEG contacts, we found that the overall sensitivity of VIES for detecting spiking SEEG contacts was 76% and its specificity for correctly identifying non-spiking SEEG contacts was 67%, indicating a good agreement between VIES and SEEG. Moreover, we found that classical dipole clustering was not informative in 9/21 patients, while VIES enable to delineate the SV in all patients. For the 12 patients having a SV delineated both with VIES and dipole clustering, VIES method had higher sensitivity and lower specificity. This proof-of-concept study shows that VIES is a promising approach to non-invasive estimation of the SV in focal epilepsy.     

Source cancellation profiles of electroencephalography and magnetoencephalography

Recorded electric potentials and magnetic fields due to cortical electrical activity have spatial spread even if their underlying brain sources are focal. Consequently, as a result of source cancellation, loss in signal amplitude and reduction in the effective signal-to-noise ratio can be expected when distributed sources are active simultaneously. Here we investigate the cancellation effects of EEG and MEG through the use of an anatomically correct forward model based on structural MRI acquired from 7 healthy adults. A boundary element model (BEM) with four compartments (brain, cerebrospinal fluid, skull and scalp) and highly accurate cortical meshes (~ 300,000 vertices) were generated. Distributed source activations were simulated using contiguous patches of active dipoles. To investigate cancellation effects in both EEG and MEG, quantitative indices were defined (source enhancement, cortical orientation disparity) and computed for varying values of the patch radius as well as for automatically parcellated gyri and sulci. Results were calculated for each cortical location, averaged over all subjects using a probabilistic atlas, and quantitatively compared between MEG and EEG. As expected, MEG sensors were found to be maximally sensitive to signals due to sources tangential to the scalp, and minimally sensitive to radial sources. Compared to EEG, however, MEG was found to be much more sensitive to signals generated antero-medially, notably in the anterior cingulate gyrus. Given that sources of activation cancel each other according to the orientation disparity of the cortex, this study provides useful methods and results for quantifying the effect of source orientation disparity upon source cancellation.     

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.     

Comparison of minimum current estimate and dipole modeling in the analysis of simulated activity in the human visual cortices

Magnetoencephalographic(MEG) data are typically interpreted using source models because of the nonunique inverse problem. Although single current dipoles, adequately representing local active areas, can be identified accurately, multiple and overlapping sources form a challenge for MEG modeling. We tested the performances of multidipole modeling and minimum current estimate (MCE) in the analysis of complicated source configurations. Simulated current sources were placed to physiologically meaningful areas of the human visual cortices. Ten volunteers from the laboratory staff analyzed four different simulations with both dipole modeling and MCE without prior information of the sources. In general, the same sources were found using both modeling methods. The subjects tended to report more false sources with MCE than with dipole model, in part due to their inexperience with the method. Dipole model was more accurate than MCE both in time and space for nonsimultaneous sources but both methods performed similarly when sources overlapped in time. For all source configurations, considerably smaller source amplitudes were reported with MCE than with dipole model.     

A spatiotemporal dynamic distributed solution to the MEG inverse problem

MEG/EEG are non-invasive imaging techniques that record brain activity with high temporal resolution. However, estimation of brain source currents from surface recordings requires solving an ill-conditioned inverse problem. Converging lines of evidence in neuroscience, from neuronal network models to resting-state imaging and neurophysiology, suggest that cortical activation is a distributed spatiotemporal dynamic process, supported by both local and long-distance neuroanatomic connections. Because spatiotemporal dynamics of this kind are central to brain physiology, inverse solutions could be improved by incorporating models of these dynamics. In this article, we present a model for cortical activity based on nearest-neighbor autoregression that incorporates local spatiotemporal interactions between distributed sources in a manner consistent with neurophysiology and neuroanatomy. We develop a dynamic Maximum a Posteriori Expectation-Maximization (dMAP-EM) source localization algorithm for estimation of cortical sources and model parameters based on the Kalman Filter, the Fixed Interval Smoother, and the EM algorithms. We apply the dMAP-EM algorithm to simulated experiments as well as to human experimental data. Furthermore, we derive expressions to relate our dynamic estimation formulas to those of standard static models, and show how dynamic methods optimally assimilate past and future data. Our results establish the feasibility of spatiotemporal dynamic estimation in large-scale distributed source spaces with several thousand source locations and hundreds of sensors, with resulting inverse solutions that provide substantial performance improvements over static methods.     

The advantage of combining MEG and EEG: comparison to fMRI in focally stimulated visual cortex

To exploit the high (millisecond) temporal resolution of magnetoencephalography (MEG) and electroencephalography (EEG) for measuring neuronal dynamics within well-defined brain regions, it is important to quantitatively assess their localizing ability. Previous modeling studies and empirical data suggest that a combination of MEG and EEG signals should yield the most accurate localization, due to their complementary sensitivities. However, these two modalities have rarely been explicitly combined for source estimation in studies of recorded brain activity, and a quantitative empirical assessment of their abilities, combined and separate, is currently lacking. Here we studied early visual responses to focal Gabor patches flashed during subject fixation. MEG and EEG data were collected simultaneously and were compared with the functional MRI (fMRI) localization produced by identical stimuli in the same subjects. This allowed direct evaluation of the localization accuracy of separate and combined MEG/EEG inverse solutions. We found that the localization accuracy of the combined MEG+EEG solution was consistently better than that of either modality alone, using three different source estimation approaches. Further analysis suggests that this improved localization is due to the different properties of the two imaging modalities rather than simply due to increased total channel number. Thus, combining MEG and EEG data is important for high-resolution spatiotemporal studies of the human brain.     

A novel integrated MEG and EEG analysis method for dipolar sources

The ability of magnetoencephalography (MEG) to accurately localize neuronal currents and obtain tangential components of the source is largely due to MEG's insensitivity to the conductivity profile of the head tissues. However, MEG cannot reliably detect the radial component of the neuronal current. In contrast, the localization accuracy of electroencephalography (EEG) is not as good as MEG, but EEG can detect both the tangential and radial components of the source. In the present study, we investigated the conductivity dependence in a new approach that combines MEG and EEG to accurately obtain, not only the location and tangential components, but also the radial component of the source. In this approach, the source location and tangential components are obtained from MEG alone, and optimal conductivity values of the EEG model are estimated by best-fitting EEG signal, while precisely matching the tangential components of the source in EEG and MEG. Then, the radial components are obtained from EEG using the previously estimated optimal conductivity values. Computer simulations testing this integrated approach demonstrated two main findings. First, there are well-organized optimal combinations of the conductivity values that provide an accurate fit to the combined MEG and EEG data. Second, the radial component, in addition to the location and tangential components, can be obtained with high accuracy without needing to know the precise conductivity profile of the head. We then demonstrated that this new approach performed reliably in an analysis of the 20-ms component from human somatosensory responses elicited by electric median-nerve stimulation.     

Reconstruction of correlated brain activity with adaptive spatial filters in MEG

Adaptive spatial filters (beamformers) have gained popularity as an effective method for the localization of brain activity from magnetoencephalography (MEG) data. Among the attractive features of some beamforming methods are high spatial resolution and no localization bias even in the presence of random noise. A drawback common to all beamforming methods, however, is significant degradation in performance in the presence of sources with high temporal correlations. Using numerical simulations and examples of auditory and visual evoked field responses, we demonstrate that, at typical signal-to-noise levels, the complete attenuation of fully correlated brain activity is less likely to occur, although significant localization and amplitude biases may occur. We compared various methods for correcting these biases and found the coherent source suppression model (CSSM) (Dalal et al., 2006) to be the most effective, with small biases for widely separated sources (e.g., bilateral auditory areas), however, amplitude biases increased systematically as distance between the sources was decreased. We assessed the performance and systematic biases that may result from the use of this model, and confirmed our findings with real examples of correlated brain activity in bilateral occipital and inferior temporal areas evoked by visually presented faces in a group of 21 adults. We demonstrated the ability to localize source activity in both regions, including correlated sources that are in close proximity (～ 3 cm) in bilateral primary visual cortex when using a priori information regarding source location. We conclude that CSSM, when carefully applied, can significantly improve localization accuracy, although amplitude biases may remain.     

A unified Bayesian framework for MEG/EEG source imaging

The ill-posed nature of the MEG (or related EEG) source localization problem requires the incorporation of prior assumptions when choosing an appropriate solution out of an infinite set of candidates. Bayesian approaches are useful in this capacity because they allow these assumptions to be explicitly quantified using postulated prior distributions. However, the means by which these priors are chosen, as well as the estimation and inference procedures that are subsequently adopted to affect localization, have led to a daunting array of algorithms with seemingly very different properties and assumptions. From the vantage point of a simple Gaussian scale mixture model with flexible covariance components, this paper analyzes and extends several broad categories of Bayesian inference directly applicable to source localization including empirical Bayesian approaches, standard MAP estimation, and multiple variational Bayesian (VB) approximations. Theoretical properties related to convergence, global and local minima, and localization bias are analyzed and fast algorithms are derived that improve upon existing methods. This perspective leads to explicit connections between many established algorithms and suggests natural extensions for handling unknown dipole orientations, extended source configurations, correlated sources, temporal smoothness, and computational expediency. Specific imaging methods elucidated under this paradigm include the weighted minimum l(2)-norm, FOCUSS, minimum current estimation, VESTAL, sLORETA, restricted maximum likelihood, covariance component estimation, beamforming, variational Bayes, the Laplace approximation, and automatic relevance determination, as well as many others. Perhaps surprisingly, all of these methods can be formulated as particular cases of covariance component estimation using different concave regularization terms and optimization rules, making general theoretical analyses and algorithmic extensions/improvements particularly relevant.     

Multiresolution imaging of MEG cortical sources using an explicit piecewise model

Imaging neural generators from MEG magnetic fields is often considered as a compromise between computationally-reasonable methodology that usually yields poor spatial resolution on the one hand, and more sophisticated approaches on the other hand, potentially leading to intractable computational costs. We approach the problem of obtaining well-resolved source images with unexcessive computation load with a multiresolution image model selection (MiMS) technique. The building blocks of the MiMS source model are parcels of the cortical surface which can be designed at multiple spatial resolutions with the combination of anatomical and functional priors. Computation charge is reduced owing to 1) compact parametric models of the activation of extended brain parcels using current multipole expansions and 2) the optimization of the generalized cross-validation error on image models, which is closed-form for the broad class of linear estimators of neural currents. Model selection can be complemented by any conventional imaging approach of neural currents restricted to the optimal image support obtained from MiMS. The estimation of the location and spatial extent of brain activations is discussed and evaluated using extensive Monte-Carlo simulations. An experimental evaluation was conducted with MEG data from a somatotopic paradigm. Results show that MiMS is an efficient image model selection technique with robust performances at realistic noise levels.