Artificial neural networks for source localization in the human brain

Electromagnetic Source Imaging via a Data-Synthesis-Based Convolutional Encoder-Decoder Network

Electromagnetic source imaging (ESI) requires solving a highly ill-posed inverse problem. To seek a unique solution, traditional ESI methods impose various forms of priors that may not accurately reflect the actual source properties, which may hinder their broad applications. To overcome this limitation, in this article, a novel data-synthesized spatiotemporally convolutional encoder-decoder network (DST-CedNet) method is proposed for ESI. The DST-CedNet recasts ESI as a machine learning problem, where discriminative learning and latent-space representations are integrated in a CedNet to learn a robust mapping from the measured electroencephalography/magnetoencephalography (E/MEG) signals to the brain activity. In particular, by incorporating prior knowledge regarding dynamical brain activities, a novel data synthesis strategy is devised to generate large-scale samples for effectively training CedNet. This stands in contrast to traditional ESI methods where the prior information is often enforced via constraints primarily aimed for mathematical convenience. Extensive numerical experiments as well as analysis of a real MEG and epilepsy EEG dataset demonstrate that the DST-CedNet outperforms several state-of-the-art ESI methods in robustly estimating source signals under a variety of source configurations.

Deep neural networks constrained by neural mass models improve electrophysiological source imaging of spatiotemporal brain dynamics

Electrophysiological source imaging (ESI) is an indispensable tool for noninvasively studying brain function and dysfunction. The moderate to low spatial resolution of electroencephalography/magnetoencephalography recordings has been a major hindrance to their wide utility in the field. Although this challenge is mitigated significantly by ESI techniques, it is difficult for individuals without relevant training to select and optimize hyperparameters for these ESI solvers. We propose a deep learning–based source imaging methodology that incorporates current advances in biophysical computational models and neural networks into the ESI framework, and it requires minimal user intervention after the model is trained. Our work promises to enable precise and robust high-resolution spatiotemporal functional brain imaging for a variety of neuroscience research studies and clinical applications.

Many efforts have been made to image the spatiotemporal electrical activity of the brain with the purpose of mapping its function and dysfunction as well as aiding the management of brain disorders. Here, we propose a non-conventional deep learning–based source imaging framework (DeepSIF) that provides robust and precise spatiotemporal estimates of underlying brain dynamics from noninvasive high-density electroencephalography (EEG) recordings. DeepSIF employs synthetic training data generated by biophysical models capable of modeling mesoscale brain dynamics. The rich characteristics of underlying brain sources are embedded in the realistic training data and implicitly learned by DeepSIF networks, avoiding complications associated with explicitly formulating and tuning priors in an optimization problem, as often is the case in conventional source imaging approaches. The performance of DeepSIF is evaluated by 1) a series of numerical experiments, 2) imaging sensory and cognitive brain responses in a total of 20 healthy subjects from three public datasets, and 3) rigorously validating DeepSIF’s capability in identifying epileptogenic regions in a cohort of 20 drug-resistant epilepsy patients by comparing DeepSIF results with invasive measurements and surgical resection outcomes. DeepSIF demonstrates robust and excellent performance, producing results that are concordant with common neuroscience knowledge about sensory and cognitive information processing as well as clinical findings about the location and extent of the epileptogenic tissue and outperforming conventional source imaging methods. The DeepSIF method, as a data-driven imaging framework, enables efficient and effective high-resolution functional imaging of spatiotemporal brain dynamics, suggesting its wide applicability and value to neuroscience research and clinical applications.

The Future of Neurotoxicology: A Neuroelectrophysiological Viewpoint

Neuroelectrophysiology is an old science, dating to the 18th century when electrical activity in nerves was discovered. Such discoveries have led to a variety of neurophysiological techniques, ranging from basic neuroscience to clinical applications. These clinical applications allow assessment of complex neurological functions such as (but not limited to) sensory perception (vision, hearing, somatosensory function), and muscle function. The ability to use similar techniques in both humans and animal models increases the ability to perform mechanistic research to investigate neurological problems. Good animal to human homology of many neurophysiological systems facilitates interpretation of data to provide cause-effect linkages to epidemiological findings. Mechanistic cellular research to screen for toxicity often includes gaps between cellular and whole animal/person neurophysiological changes, preventing understanding of the complete function of the nervous system. Building Adverse Outcome Pathways (AOPs) will allow us to begin to identify brain regions, timelines, neurotransmitters, etc. that may be Key Events (KE) in the Adverse Outcomes (AO). This requires an integrated strategy, from in vitro to in vivo (and hypothesis generation, testing, revision). Scientists need to determine intermediate levels of nervous system organization that are related to an AO and work both upstream and downstream using mechanistic approaches. Possibly more than any other organ, the brain will require networks of pathways/AOPs to allow sufficient predictive accuracy. Advancements in neurobiological techniques should be incorporated into these AOP-base neurotoxicological assessments, including interactions between many regions of the brain simultaneously. Coupled with advancements in optogenetic manipulation, complex functions of the nervous system (such as acquisition, attention, sensory perception, etc.) can be examined in real time. The integration of neurophysiological changes with changes in gene/protein expression can begin to provide the mechanistic underpinnings for biological changes. Establishment of linkages between changes in cellular physiology and those at the level of the AO will allow construction of biological pathways (AOPs) and allow development of higher throughput assays to test for changes to critical physiological circuits. To allow mechanistic/predictive toxicology of the nervous system to be protective of human populations, neuroelectrophysiology has a critical role in our future.

Review on solving the inverse problem in EEG source analysis

In this primer, we give a review of the inverse problem for EEG source localization. This is intended for the researchers new in the field to get insight in the state-of-the-art techniques used to find approximate solutions of the brain sources giving rise to a scalp potential recording. Furthermore, a review of the performance results of the different techniques is provided to compare these different inverse solutions. The authors also include the results of a Monte-Carlo analysis which they performed to compare four non parametric algorithms and hence contribute to what is presently recorded in the literature. An extensive list of references to the work of other researchers is also provided. This paper starts off with a mathematical description of the inverse problem and proceeds to discuss the two main categories of methods which were developed to solve the EEG inverse problem, mainly the non parametric and parametric methods. The main difference between the two is to whether a fixed number of dipoles is assumed a priori or not. Various techniques falling within these categories are described including minimum norm estimates and their generalizations, LORETA, sLORETA, VARETA, S-MAP, ST-MAP, Backus-Gilbert, LAURA, Shrinking LORETA FOCUSS (SLF), SSLOFO and ALF for non parametric methods and beamforming techniques, BESA, subspace techniques such as MUSIC and methods derived from it, FINES, simulated annealing and computational intelligence algorithms for parametric methods. From a review of the performance of these techniques as documented in the literature, one could conclude that in most cases the LORETA solution gives satisfactory results. In situations involving clusters of dipoles, higher resolution algorithms such as MUSIC or FINES are however preferred. Imposing reliable biophysical and psychological constraints, as done by LAURA has given superior results. The Monte-Carlo analysis performed, comparing WMN, LORETA, sLORETA and SLF, for different noise levels and different simulated source depths has shown that for single source localization, regularized sLORETA gives the best solution in terms of both localization error and ghost sources. Furthermore the computationally intensive solution given by SLF was not found to give any additional benefits under such simulated conditions.

Localization of an inert region in the brain using modified Levenberg Marquarts neural network

We tested the localization accuracy of electroencephalograph (EEG) for an inert region in a simulation at sizes ranging from 1 to 8 cm at 1 cm intervals. We used international 10-20 system electrodes placements and three concentric shell model to calculate forward problems. From using the data, neural network could be used to solve inverse problems. In this case, we estimate the localization of inert region. To demonstrate the effectiveness of the method, we perform simulations on location of inert region from EEG data, consists of training and test data. Based on the results of extensive studies, we conclude that neural network are high feasible as localization of inert region. These EEG estimation tasks were created by using a set of calculated, artificial EEG signals based on a number of current dipoles. The experimental results indicate that the proposed method has several attractive features. 1) The size of inert region is becoming more large and more the RMS values low. 2) The following the distance is closer, the RMS values is low. That could be considered inert region exists near by the electrode which has low RMS potential. 3) The more larger inert region were, the more small estimation error become.

Fast robust subject-independent magnetoencephalographic source localization using an artificial neural network

We describe a system that localizes a single dipole to reasonable accuracy from noisy magnetoencephalographic (MEG) measurements in real time. At its core is a multilayer perceptron (MLP) trained to map sensor signals and head position to dipole location. Including head position overcomes the previous need to retrain the MLP for each subject and session. The training dataset was generated by mapping randomly chosen dipoles and head positions through an analytic model and adding noise from real MEG recordings. After training, a localization took 0.7 ms with an average error of 0.90 cm. A few iterations of a Levenberg-Marquardt routine using the MLP output as its initial guess took 15 ms and improved accuracy to 0.53 cm, which approaches the natural limit on accuracy imposed by noise. We applied these methods to localize single dipole sources from MEG components isolated by blind source separation and compared the estimated locations to those generated by standard manually assisted commercial software.

MEG source localization using an MLP with a distributed output representation

We present a system that takes realistic magnetoencephalographic (MEG) signals and localizes a single dipole to reasonable accuracy in real time. At its heart is a multilayer perceptron (MLP) which takes the sensor measurements as inputs, uses one hidden layer, and generates as outputs the amplitudes of receptive fields holding a distributed representation of the dipole location. We trained this Soft-MLP on dipolar sources with real brain noise and converted the network's output into an explicit Cartesian coordinate representation of the dipole location using two different decoding strategies. The proposed Soft-MLPs are much more accurate than previous networks which output source locations in Cartesian coordinates. Hybrid Soft-MLP-start-LM systems, in which the Soft-MLP output initializes Levenberg-Marquardt, retained their accuracy of 0.28 cm with a decrease in computation time from 36 ms to 30 ms. We apply the Soft-MLP localizer to real MEG data separated by a blind source separation algorithm, and compare the Soft-MLP dipole locations to those of a conventional system.

Fast accurate MEG source localization using a multilayer perceptron trained with real brain noise

Iterative gradient methods such as Levenberg-Marquardt (LM) are in widespread use for source localization from electroencephalographic (EEG) and magnetoencephalographic (MEG) signals. Unfortunately, LM depends sensitively on the initial guess, necessitating repeated runs. This, combined with LM's high per-step cost, makes its computational burden quite high. To reduce this burden, we trained a multilayer perceptron (MLP) as a real-time localizer. We used an analytical model of quasistatic electromagnetic propagation through a spherical head to map randomly chosen dipoles to sensor activities according to the sensor geometry of a 4D Neuroimaging Neuromag-122 MEG system, and trained a MLP to invert this mapping in the absence of noise or in the presence of various sorts of noise such as white Gaussian noise, correlated noise, or real brain noise. A MLP structure was chosen to trade off computation and accuracy. This MLP was trained four times, with each type of noise. We measured the effects of initial guesses on LM performance, which motivated a hybrid MLP-start-LM method, in which the trained MLP initializes LM. We also compared the localization performance of LM, MLPs, and hybrid MLP-start-LMs for realistic brain signals. Trained MLPs are much faster than other methods, while the hybrid MLP-start-LMs are faster and more accurate than fixed-4-start-LM. In particular, the hybrid MLP-start-LM initialized by a MLP trained with the real brain noise dataset is 60 times faster and is comparable in accuracy to random-20-start-LM, and this hybrid system (localization error: 0.28 cm, computation time: 36 ms) shows almost as good performance as optimal-1-start-LM (localization error: 0.23 cm, computation time: 22 ms), which initializes LM with the correct dipole location. MLPs trained with noise perform better than the MLP trained without noise, and the MLP trained with real brain noise is almost as good an initial guesser for LM as the correct dipole location.

Electroencephalogram processing using neural networks

The electroencephalogram (EEG), a highly complex signal, is one of the most common sources of information used to study brain function and neurological disorders. More than 100 current neural network applications dedicated to EEG processing are presented. Works are categorized according to their objective (sleep analysis, monitoring anesthesia depth, brain-computer interface, EEG artifact detection, EEG source-based localization, etc.). Each application involves a specific approach (long-term analysis or short-term EEG segment analysis, real-time or time delayed processing, single or multiple EEG-channel analysis, etc.), for which neural networks were generally successful. The promising performances observed are demonstrative of the efficiency and efficacy of systems developed. This review can aid researchers, clinicians and implementors to understand up-to-date interest in neural network tools for EEG processing. The extended bibliography provides a database to assist in possible new concepts and idea development.

Estimation of intra-cranial neural activities by means of regularized neural-network-based inversion techniques

Artificial neural networks can be exploited to solve inverse problems arising from the estimation of neural activities in the brain. In this paper, we review the network inversion techniques for solving inverse problems with special attention directed towards electroencephalographic dipole localization and the improvement of positron emission tomography. In our regularized network inversion technique, for stabilizing the solution, we explicitly include the a priori knowledge by adding penalty terms to the energy function and/or build this knowledge into the architecture of the multi-layered neural networks that are used as an inverse problem solver. In the electroencephalogram analysis, the consensus term added to the energy function facilitated 3-dipole localization for visually evoked potentials. Effectiveness of our regularization is shown in improving the positron emission tomographic images and for generating metabolic images of the brain, under the constraints given by the a priori knowledge inherent to the measurement systems and physiological rules.

EEG dipole source localization using artificial neural networks

Localization of focal electrical activity in the brain using dipole source analysis of the electroencephalogram (EEG), is usually performed by iteratively determining the location and orientation of the dipole source, until optimal correspondence is reached between the dipole source and the measured potential distribution on the head. In this paper, we investigate the use of feed-forward layered artificial neural networks (ANNs) to replace the iterative localization procedure, in order to decrease the calculation time. The localization accuracy of the ANN approach is studied within spherical and realistic head models. Additionally, we investigate the robustness of both the iterative and the ANN approach by observing the influence on the localization error of both noise in the scalp potentials and scalp electrode mislocalizations. Finally, after choosing the ANN structure and size that provides a good trade off between low localization errors and short computation times, we compare the calculation times involved with both the iterative and ANN methods. An average localization error of about 3.5 mm is obtained for both spherical and realistic head models. Moreover, the ANN localization approach appears to be robust to noise and electrode mislocations. In comparison with the iterative localization, the ANN provides a major speed-up of dipole source localization. We conclude that an artificial neural network is a very suitable alternative for iterative dipole source localization in applications where large numbers of dipole localizations have to be performed, provided that an increase of the localization errors by a few millimetres is acceptable.

Comparison between human and artificial neural network detection of Laplacian-derived electroencephalographic activity related to unilateral voluntary movements

A back-propagation artificial neural network (ANN) was tested to verify its capacity to select different classes of single trials (STs) based on the spatial information content of electroencephalographic activity related to voluntary unilateral finger movements. The rationale was that ipsilateral and contralateral primary sensorimotor cortex can be involved in a nonstationary way in the control of unilateral voluntary movements. The movement-related potentials were surface Laplacian-transformed (SL) to reduce head volume conductor effects and to model the response of the primary sensorimotor cortex. The ANN sampled the SL from four or two central channels overlying the primary motor area of both sides in the period of 80 ms preceding the electromyographic response onset in the active muscle. The performance of the ANN was evaluated statistically by calculating the percentage value of agreement between the STs classified by the ANN and those of two investigators (used as a reference). The results showed that both investigator and ANN were capable of selecting STs with the SL maximum in the central area contralateral to the movement (contralateral STs, about 25%), STs with considerable SL values also in the ipsilateral central area (bilateral STs, about 50%), and STs with neither the contralateral nor bilateral pattern ("spatially incoherent" single trials; about 25%). The maximum agreement (64-84%) between the ANN and the investigator was obtained when the ANN used four spatial inputs (P < 0.0000001). Importantly, the common means of all single trials showed a weak or absent ipsilateral response. These results may suggest that a back-propagation ANN could select EEG single trials showing stationary and nonstationary responses of the primary sensorimotor cortex, based on the same spatial criteria as the experimenter.

Cardiac biomagnetic source estimation with a heart-torso model and a trained neural network

The intensity of the cardiac sources for normal adult subjects was estimated from given magnetic field profiles with a trained neural network based on the relationship of the electrical activity of the heart to the cardiac magnetic fields. The input for training the neural network consisted of the magnetic field profiles above the torso during the heartbeat. The outputs were the dipole intensities which produced those magnetic field profiles. A back propagating algorithm with bias and momentum was utilized for training. The measured and simulated torso magnetic field profiles and magnetocardiograms were used for training the neural network. Estimation of the dipole intensities was performed for unknown magnetic field profiles with the trained neural network. The estimated cardiac dipole intensities were reasonably close to the true dipole intensities. These results show the feasibility of the estimation of cardiac dipole intensities with a trained neural network under a very restricted forward model of the cardiac magnetic fields. Generalization of the results to cover a large population base could be difficult because the activation isochrones are different from subject to subject.

Variation of frontal P20 potential due to rotation of the N20-P20 dipole moment of SEPs

We have applied the artificial neural network method to estimate the N20-P20 dipole from scalp SEP potentials, and have investigated the rotation of the dipole moment caused by the compression effect of a brain tumor (para-Rolandic tumor) adjacent to the central sulcus. The variation of the frontal P20 potential was demonstrated by the rotation of the N20-P20 dipole moment in 8 cases of para-Rolandic tumor. By estimation of the rotation of the dipole moment, it may be possible to obtain preoperative information regarding the relation between the central sulcus and the tumor.

Dipole source localization of MEG by BP neural networks

The purpose of this study was to examine the usefulness of BP neural network for source localization of MEG. Since the performance of this method does not depend on the complexity of brain parameters and source models, a homogeneous brain model and a single current dipole source are assumed for convenience. Localization accuracy was examined in relation to the configuration and scale of the network. As a result, average error for position and moment estimations was within 2% while the maximum error was about 5%. It was therefore concluded that the neural network method, was useful for MEG source localization, though some improvements are still necessary.

Neural network analysis of the P300 event-related potential in multiple sclerosis

Neural network analysis is sensitive to subtle changes in patterns of data. We hypothesized that a disease process which can cause impairment of cortical function such as multiple sclerosis (MS) would affect the P300 cognitive evoked potential (P300) in a manner detectable by a feedforward backpropagation neural network. Such a network was trained using a learning data set consisting of 101 P300 wave forms (from 26 MS patients and 26 normal controls). The network was then used to classify a randomly selected test data set of 20 studies (2 studies each of 5 MS patients and 5 controls) to which it had not been previously exposed, with an average accuracy (MS = abnormal, control = normal) of 81% for a single midline electrode, increasing to 90% using 3 midline electrodes in a jury system. Neural network analysis can be of help in distinguishing normal (control) P300 from abnormal (MS) P300.