Automated MRI Volumetric Analysis in Patients with Rasmussen Syndrome.

BACKGROUND AND PURPOSE: Rasmussen syndrome, also known as Rasmussen encephalitis, is typically associated with volume loss of the affected hemisphere of the brain. Our aim was to apply automated quantitative volumetric MR imaging analyses to patients diagnosed with Rasmussen encephalitis, to determine the predictive value of lobar volumetric measures and to assess regional atrophy differences as well as monitor disease progression by using these measures. MATERIALS AND METHODS: Nineteen patients (42 scans) with diagnosed Rasmussen encephalitis were studied. We used 2 control groups: one with 42 age- and sex-matched healthy subjects and the other with 42 epileptic patients without Rasmussen encephalitis with the same disease duration as patients with Rasmussen encephalitis. Volumetric analysis was performed on T1-weighted images by using BrainSuite. Ratios of volumes from the affected hemisphere divided by those from the unaffected hemisphere were used as input to a logistic regression classifier, which was trained to discriminate patients from controls. Using the classifier, we compared the predictive accuracy of all the volumetric measures. These ratios were used to further assess regional atrophy differences and correlate with epilepsy duration. RESULTS: Interhemispheric and frontal lobe ratios had the best prediction accuracy for separating patients with Rasmussen encephalitis from healthy controls and patient controls without Rasmussen encephalitis. The insula showed significantly more atrophy compared with all the other cortical regions. Patients with longitudinal scans showed progressive volume loss in the affected hemisphere. Atrophy of the frontal lobe and insula correlated significantly with epilepsy duration. CONCLUSIONS: Automated quantitative volumetric analysis provides accurate separation of patients with Rasmussen encephalitis from healthy controls and epileptic patients without Rasmussen encephalitis, and thus may assist the diagnosis of Rasmussen encephalitis. Volumetric analysis could also be included as part of follow-up for patients with Rasmussen encephalitis to assess disease progression.

Voxel-based morphometric MRI post-processing in MRI-negative focal cortical dysplasia followed by simultaneously recorded MEG and stereo-EEG.

We aim to report on the usefulness of a voxel-based morphometric MRI post-processing technique in detecting subtle epileptogenic structural lesions. The MRI post-processing technique was implemented in a morphometric analysis program (MAP), in a 30-year-old male with pharmacoresistant focal epilepsy and negative MRI. MAP gray-white matter junction file facilitated the identification of a suspicious structural lesion in the right frontal opercular area. The electrophysiological data by simultaneously recorded stereo-EEG and MEG confirmed the epileptogenicity of the underlying subtle structural abnormality. The patient underwent a limited right frontal opercular resection, which completely included the area detected by MAP. Surgical pathology revealed focal cortical dysplasia (FCD) type IIb. Postoperatively the patient has been seizure-free for 2 years. This study demonstrates that MAP has promise in increasing the diagnostic yield of MRI reading in challenging patients with "non-lesional" MRIs. The clinical relevance and epileptogenicity of MAP abnormalities in patients with epilepsy have not been investigated systematically; therefore it is important to confirm their pertinence by performing electrophysiological recordings. When confirmed to be epileptogenic, such MAP abnormalities may reflect an underlying subtle cortical dysplasia whose complete resection can lead to seizure-free outcome.

Hidden Markov modelling of spike propagation from interictal MEG data.

For patients with partial epilepsy, automatic spike detection techniques applied to interictal MEG data often discover several potentially epileptogenic brain regions. An important determination in treatment planning is which of these detected regions are most likely to be the primary sources of epileptogenic activity. Analysis of the patterns of propagation activity between the detected regions may allow for detection of these primary epileptic foci. We describe the use of hidden Markov models (HMM) for estimation of the propagation patterns between several spiking regions from interictal MEG data. Analysis of the estimated transition probability matrix allows us to make inferences regarding the propagation pattern of the abnormal activity and determine the most likely region of its origin. The proposed HMM paradigm allows for a simple incorporation of the spike detector specificity and sensitivity characteristics. We develop bounds on performance for the case of perfect detection. We also apply the technique to simulated data sets in order to study the robustness of the method to the non-ideal specificity-sensitivity characteristics of the event detectors and compare results with the lower bounds. Our study demonstrates robustness of the proposed technique to event detection errors. We conclude with an example of the application of this method to a single patient.

On concomitant gradients in low-field MRI.

Growing interest in magnetic resonance imaging (MRI) at ultra-low magnetic fields (ULF, approximately muT fields) has been motivated by several advantages over its counterparts at higher magnetic fields. These include narrow line widths, the possibility of novel imaging schemes, reduced imaging artifacts from susceptibility variations within a sample, and reduced system cost and complexity. In addition, ULF NMR/MRI with superconducting quantum interference devices is compatible with simultaneous measurements of biomagnetic signals, a capability conventional systems cannot offer. Acquisition of MRI at ULF must, however, account for concomitant gradients that would otherwise result in severe image distortions. In this paper, we introduce the general theoretical framework that describes concomitant gradients, explain why such gradients are more problematic at low field, and present possible approaches to correct for these unavoidable gradients in the context of a non-slice-selective MRI protocol.

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.

Localization of realistic cortical activity in MEG using current multipoles.

We present a novel approach to MEG source estimation based on a regularized first-order multipole solution. The Gaussian regularizing prior is obtained by calculation of the sample mean and covariance matrix for the equivalent moments of realistic simulated cortical activity. We compare the regularized multipole localization framework to the classical dipole and general multipole source estimation methods by evaluating the ability of all three solutions to localize the centroids of physiologically plausible patches of activity simulated on the surface of a human cerebral cortex. The results, obtained with a realistic sensor configuration, a spherical head model, and given in terms of field and localization error, depict the performance of the dipolar and multipolar models as a function of variable source surface area (50-500 mm(2)), noise conditions (20, 10, and 5 dB SNR), source orientation (0-90 degrees ), and source depth (3-11 cm). We show that as the sources increase in size, they become less accurately modeled as current dipoles. The regularized multipole systematically outperforms the single dipole model, increasingly so as the spatial extent of the sources increases. In addition, our simulations demonstrate that as the orientation of the sources becomes more radial, dipole localization accuracy decreases substantially, while the performance of the regularized multipole model is far less sensitive to orientation and even succeeds in localizing quasi-radial source configurations. Furthermore, our results show that the multipole model is able to localize superficial sources with higher accuracy than the current dipole. These results indicate that the regularized multipole solution may be an attractive alternative to current-dipole-based source estimation methods in MEG.

Automated interictal spike detection and source localization in magnetoencephalography using independent components analysis and spatio-temporal clustering.

OBJECTIVE: Magnetoencephalography (MEG) dipole localization of epileptic spikes is useful in epilepsy surgery for mapping the extent of abnormal cortex and to focus intracranial electrodes. Visually analyzing large amounts of data produces fatigue and error. Most automated techniques are based on matching of interictal spike templates or predictive filtering of the data and do not explicitly include source localization as part of the analysis. This leads to poor sensitivity versus specificity characteristics. We describe a fully automated method that combines time-series analysis with source localization to detect clusters of focal neuronal current generators within the brain that produce interictal spike activity. METHODS: We first use an ICA (independent components analysis) method to decompose the multichannel MEG data and identify those components that exhibit spike-like characteristics. From these detected spikes we then find those whose spatial topographies across the array are consistent with focal neural sources, and determine the foci of equivalent current dipoles and their associated time courses. We then perform a clustering of the localized dipoles based on distance metrics that takes into consideration both their locations and time courses. The final step of refinement consists of retaining only those clusters that are statistically significant. The average locations and time series from significant clusters comprise the final output of our method. RESULTS AND SIGNIFICANCE: Data were processed from 4 patients with partial focal epilepsy. In all three subjects for whom surgical resection was performed, clusters were found in the vicinity of the resectioned area. CONCLUSIONS: The presented procedure is promising and likely to be useful to the physician as a more sensitive, automated and objective method to help in the localization of the interictal spike zone of intractable partial seizures. The final output can be visually verified by neurologists in terms of both the location and distribution of the dipole clusters and their associated time series. Due to the clinical relevance and demonstrated promise of this method, further investigation of this approach is warranted.

On MEG forward modelling using multipolar expansions.

Magnetoencephalography (MEG) is a non-invasive functional imaging modality based on the measurement of the external magnetic field produced by neural current sources within the brain. The reconstruction of the underlying sources is a severely ill-posed inverse problem typically tackled using either low-dimensional parametric source models, such as an equivalent current dipole (ECD), or high-dimensional minimum-norm imaging techniques. The inability of the ECD to properly represent non-focal sources and the over-smoothed solutions obtained by minimum-norm methods underline the need for an alternative approach. Multipole expansion methods have the advantages of the parametric approach while at the same time adequately describing sources with significant spatial extent and arbitrary activation patterns. In this paper we first present a comparative review of spherical harmonic and Cartesian multipole expansion methods that can be used in MEG. The equations are given for the general case of arbitrary conductors and realistic sensor configurations and also for the special cases of spherically symmetric conductors and radially oriented sensors. We then report the results of computer simulations used to investigate the ability of a first-order multipole model (dipole and quadrupole) to represent spatially extended sources, which are simulated by 2D and 3D clusters of elemental dipoles. The overall field of a cluster is analysed using singular value decomposition and compared to the unit fields of a multipole, centred in the middle of the cluster, using subspace correlation metrics. Our results demonstrate the superior utility of the multipolar source model over ECD models in providing source representations of extended regions of activity.

Rapidly recomputable EEG forward models for realistic head shapes.

With the increasing availability of surface extraction techniques for magnetic resonance and x-ray computed tomography images, realistic head models can be readily generated as forward models in the analysis of electroencephalography (EEG) and magnetoencephalography (MEG) data. Inverse analysis of this data, however, requires that the forward model be computationally efficient. We propose two methods for approximating the EEG forward model using realistic head shapes. The 'sensor-fitted sphere' approach fits a multilayer sphere individually to each sensor, and the 'three-dimensional interpolation' scheme interpolates using a grid on which a numerical boundary element method (BEM) solution has been precomputed. We have characterized the performance of each method in terms of magnitude and subspace error metrics, as well as computational and memory requirements. We have also made direct performance comparisons with traditional spherical models. The approximation provided by the interpolative scheme had an accuracy nearly identical to full BEM, even within 3 mm of the inner skull surface. Forward model computation during inverse procedures was approximately 30 times faster than for a traditional three-shell spherical model. Cast in this framework, high-fidelity numerical solutions currently viewed as computationally prohibitive for solving the inverse problem (e.g. linear Galerkin BEM) can be rapidly recomputed in a highly efficient manner. The sensor-fitting method has a similar one-time cost to the BEM method, and while it produces some improvement over a standard three-shell sphere, its performance does not approach that of the interpolation method. In both methods, there is a one-time cost associated with precomputing the forward solution over a set of grid points.

Paired MEG data set source localization using recursively applied and projected (RAP) MUSIC.

An important class of experiments in functional brain mapping involves collecting pairs of data corresponding to separate "Task" and "Control" conditions. The data are then analyzed to determine what activity occurs during the Task experiment but not in the Control. Here we describe a new method for processing paired magnetoencephalographic (MEG) data sets using our recursively applied and projected multiple signal classification (RAP-MUSIC) algorithm. In this method the signal subspace of the Task data is projected against the orthogonal complement of the Control data signal subspace to obtain a subspace which describes spatial activity unique to the Task. A RAP-MUSIC localization search is then performed on this projected data to localize the sources which are active in the Task but not in the Control data. In addition to dipolar sources, effective blocking of more complex sources, e.g., multiple synchronously activated dipoles or synchronously activated distributed source activity, is possible since these topographies are well-described by the Control data signal subspace. Unlike previously published methods, the proposed method is shown to be effective in situations where the time series associated with Control and Task activity possess significant cross correlation. The method also allows for straightforward determination of the estimated time series of the localized target sources. A multiepoch MEG simulation and a phantom experiment are presented to demonstrate the ability of this method to successfully identify sources and their time series in the Task data.

EEG source localization and imaging using multiple signal classification approaches.

Equivalent current dipoles are a powerful tool for modeling focal sources. The dipole is often sufficient to adequately represent sources of measured scalp potentials, even when the area of activation exceeds 1 cm2 of cortex. Traditional least-squares fitting techniques involve minimization of an error function with respect to the location and orientation of the dipoles. The existence of multiple local minima in this error function can result in gross errors in the computed source locations. The problem is further compounded by the requirement that the model order, i.e. the number of dipoles, be determined before error minimization can be performed. An incorrect model order can produce additional errors in the estimated source parameters. Both of these problems can be avoided using alternative search strategies based on the MUSIC (multiple signal classification) algorithm. Here the authors review the MUSIC approach and demonstrate its application to the localization of multiple current dipoles from EEG data. The authors also show that the number of detectable sources can be determined in a recursive manner from the data. Also, in contrast to least-squares, the method can find dipolar sources in the presence of additional non-dipolar sources. Finally, extensions of the MUSIC approach to allow the modeling of distributed sources are discussed.

EEG and MEG: forward solutions for inverse methods.

A solution of the forward problem is an important component of any method for computing the spatio-temporal activity of the neural sources of magnetoencephalography (MEG) and electroencephalography (EEG) data. The forward problem involves computing the scalp potentials or external magnetic field at a finite set of sensor locations for a putative source configuration. We present a unified treatment of analytical and numerical solutions of the forward problem in a form suitable for use in inverse methods. This formulation is achieved through factorization of the lead field into the product of the moment of the elemental current dipole source with a "kernel matrix" that depends on the head geometry and source and sensor locations, and a "sensor matrix" that models sensor orientation and gradiometer effects in MEG and differential measurements in EEG. Using this formulation and a recently developed approximation formula for EEG, based on the "Berg parameters," we present novel reformulations of the basic EEG and MEG kernels that dispel the myth that EEG is inherently more complicated to calculate than MEG. We also present novel investigations of different boundary element methods (BEM's) and present evidence that improvements over currently published BEM methods can be realized using alternative error-weighting methods. Explicit expressions for the matrix kernels for MEG and EEG for spherical and realistic head geometries are included.

A sensor-weighted overlapping-sphere head model and exhaustive head model comparison for MEG.

The spherical head model has been used in magnetoencephalography (MEG) as a simple forward model for calculating the external magnetic fields resulting from neural activity. For more realistic head shapes, the boundary element method (BEM) or similar numerical methods are used, but at greatly increased computational cost. We introduce a sensor-weighted overlapping-sphere (OS) head model for rapid calculation of more realistic head shapes. The volume currents associated with primary neural activity are used to fit spherical head models for each individual MEG sensor such that the head is more realistically modelled as a set of overlapping spheres, rather than a single sphere. To assist in the evaluation of this OS model with BEM and other head models, we also introduce a novel comparison technique that is based on a generalized eigenvalue decomposition and accounts for the presence of noise in the MEG data. With this technique we can examine the worst possible errors for thousands of dipole locations in a realistic brain volume. We test the traditional single-sphere model, three-shell and single-shell BEM, and the new OS model. The results show that the OS model has accuracy similar to the BEM but is orders of magnitude faster to compute.

Recursive MUSIC: a framework for EEG and MEG source localization.

The multiple signal classification (MUSIC) algorithm can be used to locate multiple asynchronous dipolar sources from electroencephalography (EEG) and magnetoencephalography (MEG) data. The algorithm scans a single-dipole model through a three-dimensional (3-D) head volume and computes projections onto an estimated signal subspace. To locate the sources, the user must search the head volume for multiple local peaks in the projection metric. This task is time consuming and subjective. Here, we describe an extension of this approach which we refer to as recursive MUSIC (R-MUSIC). This new procedure automatically extracts the locations of the sources through a recursive use of subspace projections. The new method is also able to locate synchronous sources through the use of a spatio-temporal independent topographies (IT) model. This model defines a source as one or more nonrotating dipoles with a single time course. Within this framework, we are able to locate fixed, rotating, and synchronous dipoles. The recursive subspace projection procedure that we introduce here uses the metric of canonical or subspace correlations as a multidimensional form of correlation analysis between the model subspace and the data subspace. By recursively computing subspace correlations, we build up a model for the sources which account for a given set of data. We demonstrate here how R-MUSIC can easily extract multiple asynchronous dipolar sources that are difficult to find using the original MUSIC scan. We then demonstrate R-MUSIC applied to the more general IT model and show results for combinations of fixed, rotating, and synchronous dipoles.

A study of dipole localization accuracy for MEG and EEG using a human skull phantom.

OBJECTIVE: To investigate the accuracy of forward and inverse techniques for EEG and MEG dipole localization. DESIGN AND METHODS: A human skull phantom was constructed with brain, skull and scalp layers and realistic relative conductivities. Thirty two independent current dipoles were distributed within the 'brain' region and EEG and MEG data collected separately for each dipole. The true dipole locations and orientations and the morphology of the brain, skull and scalp layers were extracted from X-ray CT data. The location of each dipole was estimated from the EEG and MEG data using the R-MUSIC inverse method and forward models based on spherical and realistic head geometries. Additional computer simulations were performed to investigate the factors affecting localization accuracy. RESULTS: Localization errors using the relatively simpler locally fitted sphere approach are only slightly greater than those using a BEM approach. The average localization error over the 32 dipoles was 7-8 mm for EEG and 3 mm for MEG. CONCLUSION: The superior performance of MEG over EEG appears to be because the latter is more sensitive to errors in the forward model arising from simplifying assumptions concerning the conductivity of the skull, scalp and brain.

MEG-based imaging of focal neuronal current sources.

We describe a new approach to imaging neural current sources from measurements of the magnetoencephalogram (MEG) associated with sensory, motor, or cognitive brain activation. Many previous approaches to this problem have concentrated on the use of weighted minimum norm (WMN) inverse methods. While these methods ensure a unique solution, they do not introduce information specific to the MEG inverse problem, often producing overly smoothed solutions and exhibiting severe sensitivity to noise. We describe a Bayesian formulation of the inverse problem in which a Gibbs prior is constructed to reflect the sparse focal nature of neural current sources associated with evoked response data. We demonstrate the method with simulated and experimental phantom data, comparing its performance with several WMN methods.

Mapping function in the human brain with magnetoencephalography, anatomical magnetic resonance imaging, and functional magnetic resonance imaging.

Integrated analyses of human anatomical and functional measurements offer a powerful paradigm for human brain mapping. Magnetoencephalography (MEG) and EEG provide excellent temporal resolution of neural population dynamics as well as capabilities for source localization. Anatomical magnetic resonance imaging (MRI) provides excellent spatial resolution of head and brain anatomy, whereas functional MRI (fMRI) techniques provide an alternative measure of neural activation based on associated hemodynamic changes. These methodologies constrain and complement each other and can thereby improve our interpretation of functional neural organization. We have developed a number of computational tools and techniques for the visualization, comparison, and integrated analysis of multiple neuroimaging techniques. Construction of geometric anatomical models from volumetric MRI data allows improved models of the head volume conductor and can provide powerful constraints for neural electromagnetic source modeling. These approaches, coupled to enhanced algorithmic strategies for the inverse problem, can significantly enhance the accuracy of source-localization procedures. We have begun to apply these techniques for studies of the functional organization of the human visual system. Such studies have demonstrated multiple, functionally distinct visual areas that can be resolved on the basis of their locations, temporal dynamics, and differential sensitivity to stimulus parameters. Our studies have also produced evidence of internal retinotopic organization in both striate and extrastriate visual areas but have disclosed organizational departures from classical models. Comparative studies of MEG and fMRI suggest a reasonable but imperfect correlation between electrophysiological and hemodynamic responses. We have demonstrated a method for the integrated analysis of fMRI and MEG, and we outline strategies for improvement of these methods. By combining multiple measurement techniques, we can exploit the complementary strengths and transcend the limitations of the individual neuro-imaging methods.

Error bounds for EEG and MEG dipole source localization.

General formulas are presented for computing a lower bound on localization and moment error for electroencephalographic (EEG) or magnetoencephalographic (MEG) current source dipole models with arbitrary sensor array geometry. Specific EEG and MEG formulas are presented for multiple dipoles in a head model with 4 spherical shells. Localization error bounds are presented for both EEG and MEG for several different sensor configurations. Graphical error contours are presented for 127 sensors covering the upper hemisphere, for both 37 sensors and 127 sensors covering a smaller region, and for the standard 10-20 EEG sensor arrangement. Both 1- and 2-dipole cases were examined for all possible dipole orientations and locations within a head quadrant. The results show a strong dependence on absolute dipole location and orientation. The results also show that fusion of the EEG and MEG measurements into a combined model reduces the lower bound. A Monte Carlo simulation was performed to check the tightness of the bounds for a selected case. The simple head model, the low power noise and the few strong dipoles were all selected in this study as optimistic conditions to establish possibly fundamental resolution limits for any localization effort. Results, under these favorable assumptions, show comparable resolutions between the EEG and the MEG models, but accuracy for a single dipole, in either case, appears limited to several millimeters for a single time slice. The lower bounds increase markedly with just 2 dipoles. Observations are given to support the need for full spatiotemporal modeling to improve these lower bounds. All of the simulation results presented can easily be scaled to other instances of noise power and dipole intensity.

Multiple dipole modeling and localization from spatio-temporal MEG data.

An array of biomagnetometers may be used to measure the spatio-temporal neuromagnetic field or magnetoencephalogram (MEG) produced by neural activity in the brain. A popular model for the neural activity produced in response to a given sensory stimulus is a set of current dipoles, where each dipole represents the primary current associated with the combined activation of a large number of neurons located in a small volume of the brain. An important problem in the interpretation of MEG data from evoked response experiments is the localization of these neural current dipoles. We present here a linear algebraic framework for three common spatio-temporal dipole models: i) unconstrained dipoles, ii) dipoles with a fixed location, and iii) dipoles with a fixed orientation and location. In all cases, we assume that the location, orientation, and magnitude of the dipoles are unknown. With a common model, we show how the parameter estimation problem may be decomposed into the estimation of the time invariant parameters using nonlinear least-squares minimization, followed by linear estimation of the associated time varying parameters. A subspace formulation is presented and used to derive a suboptimal least-squares subspace scanning method. The resulting algorithm is a special case of the well-known MUltiple SIgnal Classification (MUSIC) method, in which the solution (multiple dipole locations) is found by scanning potential locations using a simple one dipole model. Principal components analysis (PCA) dipole fitting has also been used to individually fit single dipoles in a multiple dipole problem. Analysis is presented here to show why PCA dipole fitting will fail in general, whereas the subspace method presented here will generally succeed. Numerically efficient means of calculating the cost functions are presented, and problems of model order selection and missing moments are discussed. Results from a simulation and a somatosensory experiment are presented.