EEG Dipole Source Localization using Deep Neural Network

Electroencephalography (EEG) inverse problem is a typical inverse problem, in which the electrical activity within the brain is reconstructed based on EEG data collected from the scalp electrodes. In this paper, the four-layer concentric head model is used for simulation firstly, four deep neural network models including a multilayer perceptron (MLP) model and three convolutional neural networks (CNNs) are adopted to solve EEG inverse problem based on equal current dipole (ECD) model. In the simulations, 100,000 samples are generated randomly, of which 60% are used for network training and 20% are used for cross-validation. Eventually, the generalization performance of the model using the optimal function is measured by the errors in the rest 20% testing set. The experimental results show that the absolute error, relative error, mean positioning error and standard deviation of the four models are extremely low. The CNN with 6 convolutional layers and 3 pooling layers (CNN-3) is the best model. Its absolute error is about 0.015, its relative error is about 0.005, and its dipole position error is 0.040±0.029 cm. Furthermore, we use CNN-3 for source localization of the real EEG data in Working Memory. The results are in accord with physiological experience. The deep neural network method in our study needs fewer calculation parameters, takes less time, and has better positioning results.

Findings about LORETA Applied to High-Density EEG—A Review

Electroencephalography (EEG) is a non-invasive diagnostic technique for recording brain electric activity. The EEG source localization has been an area of research widely explored during the last decades because it provides helpful information about brain physiology and abnormalities. Source localization consists in solving the so-called EEG inverse problem. Over the years, one of the most employed method for solving it has been LORETA (Low Resolution Electromagnetic Tomography). In particular, in this review, we focused on the findings about the LORETA family algorithms applied to high-density EEGs (HD-EEGs), used for improving the low spatial resolution deriving from the traditional EEG systems. The results were classified according to their clinical application and some aspects arisen from the analyzed papers were discussed. Finally, suggestions were provided for future improvement. In this way, the combination of LORETA with HD-EEGs could become an even more valuable tool for noninvasive clinical evaluation in the field of applied neuroscience.

Electrical Dipole Source Localization using Hybrid Least Squares Method in combination with ICA

In biomedical engineering, dipole source localization is commonly used to identify brain activities from scalp recorded potentials, which is known as inverse problem of electroencephalography (EEG) source localization. However, this problem is fundamental in biomedical engineering, medicine and neuroscience. The EEG inverse problem is non-linear, in addition, it is ill-posed and the solver can be unstable, i.e. the solution is non-unique and it is highly sensitive to small changes of the measured signal (noise). For solving the EEG inverse problem iterative methods, like Levenberg-Marquardt algorithm, are usually considered. However, these techniques require good initial values and many electrodes N, since a large redundancy supports the finding of the right solution. Therefore, in this paper, a hybrid method of linear and non-linear modelling and least squares approach are proposed to overcome of these problems: the solutions calculated by means of a linear approximation of EEG inverse problems serve as initial values for solving the original non-linear model. In addition, independent component analysis (ICA) is combined with the proposed hybrid least squares method to separate different dipole sources from multiple EEG signals. The performance of the hybrid least squares method with and without ICA is measured in term of root mean square error. The simulation results show that the proposed method can estimate the location of dipole source with acceptable accuracy under high noise condition and small N comparing with linear least squares method considering larger N. Finally, it should be mentioned that the proposed method promises advantages in finding solutions of the EEG inverse problem effectively.