Search

To reconstruct the electrophysiological activity of brain responses, source analysis is performed through the solution of the forward and inverse problems. The former contains a unique solution while the latter is ill-posed. In this regard, many algorithms have been suggested relying on different prior information for solving the inverse problem. Recently, neural networks have been used to deal with source analysis. However, their underlying training for inverse solutions is based on suboptimal forward modeling. In this work, we propose a CNN that is able to reconstruct EEG brain activity. To train our proposed CNN, a skull-conductivity calibrated and white matter anisotropic head model. Based on this model, we generate simulated EEG data and used them to train our CNN. We first evaluate the performance of our CNN using the simulated EEG data while a realistic application with somatosensory evoked potentials follows. From the results, we observed that the CCN correctly localized the P20/N20 component at the subject-specific Brodmann area 3b and it can potentially localize deeper sources. A comparison is also presented with well-known inverse solutions (single dipole scans and sLORETA) showing similar localization performance. Through these results, an emerging potential for real applications appears on the basis of realistic head modeling.

The aim of this paper is to advance electroencephalography (EEG) source analysis using finite element method (FEM) head volume conductor models that go beyond the standard three compartment (skin, skull, brain) approach and take brain tissue inhomogeneity (gray and white matter and cerebrospinal fluid) into account. The new approach should enable accurate EEG forward modeling in the thin human cortical structures and, more specifically, in the especially thin cortices in children brain research or in pathological applications. The source model should thus be focal enough to be usable in the thin cortices, but should on the other side be more realistic than the current standard mathematical point dipole. Furthermore, it should be numerically accurate and computationally fast. We propose to achieve the best balance between these demands with a current preserving (divergence conforming) dipolar source model. We develop and investigate a varying number of current preserving source basis elements n (n=1,…,n=5). For validation, we conducted numerical experiments within a multi-layered spherical domain, where an analytical solution exists. We show that the accuracy increases along with the number of basis elements, while focality decreases. The results suggest that the best balance between accuracy and focality in thin cortices is achieved with n=4 (or in extreme cases even n=3) basis functions, while in thicker cortices n=5 is recommended to obtain the highest accuracy. We also compare the current preserving approach to two further FEM source modeling techniques, namely partial integration and St. Venant, and show that the best current preserving source model outperforms the competing methods with regard to overall balance. For all tested approaches, FEM transfer matrices enable high computational speed. We implemented the new EEG forward modeling approaches into the open source duneuro library for forward modeling in bioelectromagnetism to enable its broader use by the brain research community. This library is build upon the DUNE framework for parallel finite elements simulations and integrates with high-level toolboxes like FieldTrip. Additionally, an inversion test has been implemented using the realistic head model to demonstrate and compare the differences between the aforementioned source models., (Copyright © 2018 Elsevier Inc. All rights reserved.)

The goal of this study is to develop focal, accurate and robust finite element method (FEM) based approaches which can predict the electric potential on the surface of the computational domain given its structure and internal primary source current distribution. While conducting an EEG evaluation, the placement of source currents to the geometrically complex grey matter compartment is a challenging but necessary task to avoid forward errors attributable to tissue conductivity jumps. Here, this task is approached via a mathematically rigorous formulation, in which the current field is modeled via divergence conforming H(div) basis functions. Both linear and quadratic functions are used while the potential field is discretized via the standard linear Lagrangian (nodal) basis. The resulting model includes dipolar sources which are interpolated into a random set of positions and orientations utilizing two alternative approaches: the position based optimization (PBO) and the mean position/orientation (MPO) method. These results demonstrate that the present dipolar approach can reach or even surpass, at least in some respects, the accuracy of two classical reference methods, the partial integration (PI) and St. Venant (SV) approach which utilize monopolar loads instead of dipolar currents.

To explore the relationship between transcranial current stimulation (tCS) and the electroencephalography (EEG) forward problem, we investigate and compare accuracy and efficiency of a reciprocal and a direct EEG forward approach for dipolar primary current sources both based on the finite element method (FEM), namely the adjoint approach (AA) and the partial integration approach in conjunction with a transfer matrix concept (PI). By analyzing numerical results, comparing to analytically derived EEG forward potentials and estimating computational complexity in spherical shell models, AA turns out to be essentially identical to PI. It is then proven that AA and PI are also algebraically identical even for general head models. This relation offers a direct link between the EEG forward problem and tCS. We then demonstrate how the quasi-analytical EEG forward solutions in sphere models can be used to validate the numerical accuracies of FEM-based tCS simulation approaches. These approaches differ with respect to the ease with which they can be employed for realistic head modeling based on MRI-derived segmentations. We show that while the accuracy of the most easy to realize approach based on regular hexahedral elements is already quite high, it can be significantly improved if a geometry-adaptation of the elements is employed in conjunction with an isoparametric FEM approach. While the latter approach does not involve any additional difficulties for the user, it reaches the high accuracies of surface-segmentation based tetrahedral FEM, which is considerably more difficult to implement and topologically less flexible in practice. Finally, in a highly realistic head volume conductor model and when compared to the regular alternative, the geometry-adapted hexahedral FEM is shown to result in significant changes in tCS current flow orientation and magnitude up to 45° and a factor of 1.66, respectively., (Copyright © 2016 Elsevier Inc. All rights reserved.)

The high temporal resolution of EEG/MEG data offers a way to improve source reconstruction estimates which provide insight into the spatio-temporal involvement of neuronal sources in the human brain. In this work, we investigated the performance of spatio-temporal regularization (STR) in a current density approach using a systematic comparison to simple ad hoc or post hoc filtering of the data or of the reconstructed current density, respectively. For the used STR approach we implemented a frequency-specific constraint to penalize solutions outside a narrow frequency band of interest. The widely used sLORETA algorithm was adapted for STR and generally used for source reconstruction. STR and filtering approaches were evaluated with respect to spatial localization error and spatial dispersion, as well as to correlation of original and reconstructed source time courses in single source and two source scenarios with fixed source locations and oscillating source waveforms. We used extensive computer simulations and tested all algorithms with different parameter settings (noise levels and regularization parameters) for EEG data. To verify our results, we also used data from MEG phantom measurements. For the investigated scenarios, we did not find any evidence that STR-based methods outperform purely spatial algorithms applied to temporally filtered data. Furthermore, the results show very clearly that the performance of STR depends very much on the choice of regularization parameters.

The low-conducting human skull is known to have an especially large influence on electroencephalography (EEG) source analysis. Because of difficulties segmenting the complex skull geometry out of magnetic resonance images, volume conductor models for EEG source analysis might contain inaccuracies and simplifications regarding the geometry of the skull. The computer simulation study presented here investigated the influences of a variety of skull geometry deficiencies on EEG forward simulations and source reconstruction from EEG data. Reference EEG data was simulated in a detailed and anatomically plausible reference model. Test models were derived from the reference model representing a variety of skull geometry inaccuracies and simplifications. These included erroneous skull holes, local errors in skull thickness, modeling cavities as bone, downward extension of the model and simplifying the inferior skull or the inferior skull and scalp as layers of constant thickness. The reference EEG data was compared to forward simulations in the test models, and source reconstruction in the test models was performed on the simulated reference data. The finite element method with high-resolution meshes was employed for all forward simulations. It was found that large skull geometry inaccuracies close to the source space, for example, when cutting the model directly below the skull, led to errors of 20mm and more for extended source space regions. Local defects, for example, erroneous skull holes, caused non-negligible errors only in the vicinity of the defect. The study design allowed a comparison of influence size, and guidelines for modeling the skull geometry were concluded., (Copyright © 2012 Elsevier Inc. All rights reserved.)

The estimation of the activity-related ion currents by measuring the induced electromagnetic fields at the head surface is a challenging and severely ill-posed inverse problem. This is especially true in the recovery of brain networks involving deep-lying sources by means of EEG/MEG recordings which is still a challenging task for any inverse method. Recently, hierarchical Bayesian modeling (HBM) emerged as a unifying framework for current density reconstruction (CDR) approaches comprising most established methods as well as offering promising new methods. Our work examines the performance of fully-Bayesian inference methods for HBM for source configurations consisting of few, focal sources when used with realistic, high-resolution finite element (FE) head models. The main foci of interest are the correct depth localization, a well-known source of systematic error of many CDR methods, and the separation of single sources in multiple-source scenarios. Both aspects are very important in the analysis of neurophysiological data and in clinical applications. For these tasks, HBM provides a promising framework and is able to improve upon established CDR methods such as minimum norm estimation (MNE) or sLORETA in many aspects. For challenging multiple-source scenarios where the established methods show crucial errors, promising results are attained. Additionally, we introduce Wasserstein distances as performance measures for the validation of inverse methods in complex source scenarios., (Copyright © 2012 Elsevier Inc. All rights reserved.)

We used computer simulations to investigate finite element models of the layered structure of the human skull in EEG source analysis. Local models, where each skull location was modeled differently, and global models, where the skull was assumed to be homogeneous, were compared to a reference model, in which spongy and compact bone were explicitly accounted for. In both cases, isotropic and anisotropic conductivity assumptions were taken into account. We considered sources in the entire brain and determined errors both in the forward calculation and the reconstructed dipole position. Our results show that accounting for the local variations over the skull surface is important, whereas assuming isotropic or anisotropic skull conductivity has little influence. Moreover, we showed that, if using an isotropic and homogeneous skull model, the ratio between skin/brain and skull conductivities should be considerably lower than the commonly used 80:1. For skull modeling, we recommend (1) Local models: if compact and spongy bone can be identified with sufficient accuracy (e.g., from MRI) and their conductivities can be assumed to be known (e.g., from measurements), one should model these explicitly by assigning each voxel to one of the two conductivities, (2) Global models: if the conditions of (1) are not met, one should model the skull as either homogeneous and isotropic, but with considerably higher skull conductivity than the usual 0.0042 S/m, or as homogeneous and anisotropic, but with higher radial skull conductivity than the usual 0.0042 S/m and a considerably lower radial:tangential conductivity anisotropy than the usual 1:10., (Copyright © 2010 Wiley-Liss, Inc.)

Determining the centers of electrical activity in the human body and the connectivity between different centers of activity in the brain is an active area of research. To understand brain function and the nature of cardiovascular diseases requires sophisticated methods applicable to non-invasively measured bioelectric and biomagnetic data. As it is difficult to solve for all unknown parameters at once, several strains of data analysis have been developed, each trying to solve a different part of the problem and each requiring a different set of assumptions. Current trends and results from major topics of electro- and magnetoencephalographic data analysis are presented here together with the aim of stimulating research into the unification of the different approaches. The following topics are discussed: source reconstruction using detailed finite element modeling to locate sources deep in the brain; connectivity analysis for the quantification of strength and direction of information flow between activity centers, preferably incorporating an inverse solution; the conflict between the statistical independence assumption of sources and a possible connectivity; the verification and validation of results derived from non-invasively measured data through animal studies and phantom measurements. This list already indicates the benefits of a unified view.

Bioelectric source analysis in the human brain from scalp electroencephalography (EEG) signals is sensitive to geometry and conductivity properties of the different head tissues. We propose a low-resolution conductivity estimation (LRCE) method using simulated annealing optimization on high-resolution finite element models that individually optimizes a realistically shaped four-layer volume conductor with regard to the brain and skull compartment conductivities. As input data, the method needs T1- and PD-weighted magnetic resonance images for an improved modeling of the skull and the cerebrospinal fluid compartment and evoked potential data with high signal-to-noise ratio (SNR). Our simulation studies showed that for EEG data with realistic SNR, the LRCE method was able to simultaneously reconstruct both the brain and the skull conductivity together with the underlying dipole source and provided an improved source analysis result. We have also demonstrated the feasibility and applicability of the new method to simultaneously estimate brain and skull conductivity and a somatosensory source from measured tactile somatosensory-evoked potentials of a human subject. Our results show the viability of an approach that computes its own conductivity values and thus reduces the dependence on assigning values from the literature and likely produces a more robust estimate of current sources. Using the LRCE method, the individually optimized four-compartment volume conductor model can, in a second step, be used for the analysis of clinical or cognitive data acquired from the same subject., (2008 Wiley-Liss, Inc.)

The major goal of the evaluation in presurgical epilepsy diagnosis for medically intractable patients is the precise reconstruction of the epileptogenic foci, preferably with non-invasive methods. This paper evaluates whether surface electroencephalography (EEG) source analysis based on a 1 mm anisotropic finite element (FE) head model can provide additional guidance for presurgical epilepsy diagnosis and whether it is practically feasible in daily routine. A 1 mm hexahedra FE volume conductor model of the patient's head with special focus on accurately modeling the compartments skull, cerebrospinal fluid (CSF) and the anisotropic conducting brain tissues was constructed using non-linearly co-registered T1-, T2- and diffusion-tensor-magnetic resonance imaging data. The electrodes of intra-cranial EEG (iEEG) measurements were extracted from a co-registered computed tomography image. Goal function scan (GFS), minimum norm least squares (MNLS), standardized low resolution electromagnetic tomography (sLORETA) and spatio-temporal current dipole modeling inverse methods were then applied to the peak of the averaged ictal discharges EEG data. MNLS and sLORETA pointed to a single center of activity. Moving and rotating single dipole fits resulted in an explained variance of more than 97%. The non-invasive EEG source analysis methods localized at the border of the lesion and at the border of the iEEG electrodes which mainly received ictal discharges. Source orientation was towards the epileptogenic tissue. For the reconstructed superficial source, brain conductivity anisotropy and the lesion conductivity had only a minor influence, whereas a correct modeling of the highly conducting CSF compartment and the anisotropic skull was found to be important. The proposed FE forward modeling approach strongly simplifies meshing and reduces run-time (37 ms for one forward computation in the model with 3.1 million unknowns), corroborating the practical feasibility of the approach.

Mesh generation in finite-element- (FE) method-based electroencephalography (EEG) source analysis generally influences greatly the accuracy of the results. It is thus important to determine a meshing strategy well adopted to achieve both acceptable accuracy for potential distributions and reasonable computation times and memory usage. In this paper, we propose to achieve this goal by smoothing regular hexahedral finite elements at material interfaces using a node-shift approach. We first present the underlying theory for two different techniques for modeling a current dipole in FE volume conductors, a subtraction and a direct potential method. We then evaluate regular and smoothed elements in a four-layer sphere model for both potential approaches and compare their accuracy. We finally compute and visualize potential distributions for a tangentially and a radially oriented source in the somatosensory cortex in regular and geometry-adapted three-compartment hexahedra FE volume conductor models of the human head using both the subtraction and the direct potential method. On the average, node-shifting reduces both topography and magnitude errors by more than a factor of 2 for tangential and 1.5 for radial sources for both potential approaches. Nevertheless, node-shifting has to be carried out with caution for sources located within or close to irregular hexahedra, because especially for the subtraction method extreme deformations might lead to larger overall errors. With regard to realistic volume conductor modeling, node-shifted hexahedra should thus be used for the skin and skull compartments while we would not recommend deforming elements at the grey and white matter surfaces.

The aim of our work was to quantify the influence of white matter anisotropic conductivity information on electroencephalography (EEG) source reconstruction. We performed this quantification in a rabbit head using both simulations and source localization based on invasive measurements. In vivo anisotropic (tensorial) conductivity information was obtained from magnetic resonance diffusion tensor imaging and included into a high-resolution finite-element model. When neglecting anisotropy in the simulations, we found a shift in source location of up to 1.3 mm with a mean value of 0.3 mm. The averaged orientational deviation was 10 degree and the mean magnitude error of the dipole was 29%. Source localization of the first cortical components after median and tibial nerve stimulation resulted in anatomically verified dipole positions with no significant anisotropy effect. Our results indicate that the expected average source localization error due to anisotropic white matter conductivity is within the principal accuracy limits of current inverse procedures. However, larger localization errors might occur in certain cases. In contrast, dipole orientation and dipole strength are influenced significantly by the anisotropy. We conclude that the inclusion of tissue anisotropy information improves source estimation procedures.