Sung Chan Jun

Publications

• Scanning Reduction Strategy in MEG/EEG Beamformer Source Imaging.

  Jun Hee Hong, Sung Chan Jun

  J. Applied Mathematics. 01/2012; 2012.
• 2.30

  Impact points

  Feasibility of approaches combining sensor and source features in brain-computer interface.

  Minkyu Ahn, Jun Hee Hong, Sung Chan Jun

  Journal of neuroscience methods. 11/2011; 204(1):168-78.

  Brain-computer interface (BCI) provides a new channel for communication between brain and computers through brain signals. Cost-effective EEG provides good temporal resolution, but its spatial resolution is poor and sensor information is blurred by inherent noise. To overcome these issues, spatial f... [more] Brain-computer interface (BCI) provides a new channel for communication between brain and computers through brain signals. Cost-effective EEG provides good temporal resolution, but its spatial resolution is poor and sensor information is blurred by inherent noise. To overcome these issues, spatial filtering and feature extraction techniques have been developed. Source imaging, transformation of sensor signals into the source space through source localizer, has gained attention as a new approach for BCI. It has been reported that the source imaging yields some improvement of BCI performance. However, there exists no thorough investigation on how source imaging information overlaps with, and is complementary to, sensor information. Information (visible information) from the source space may overlap as well as be exclusive to information from the sensor space is hypothesized. Therefore, we can extract more information from the sensor and source spaces if our hypothesis is true, thereby contributing to more accurate BCI systems. In this work, features from each space (sensor or source), and two strategies combining sensor and source features are assessed. The information distribution among the sensor, source, and combined spaces is discussed through a Venn diagram for 18 motor imagery datasets. Additional 5 motor imagery datasets from the BCI Competition III site were examined. The results showed that the addition of source information yielded about 3.8% classification improvement for 18 motor imagery datasets and showed an average accuracy of 75.56% for BCI Competition data. Our proposed approach is promising, and improved performance may be possible with better head model.
• 1.43

  Impact points

  Validation of weighted frequency-difference EIT using a three-dimensional hemisphere model and phantom.

  Sujin Ahn, Tong In Oh, Sung Chan Jun, Jin Keun Seo, Eung Je Woo

  Physiological measurement. 09/2011; 32(10):1663-80.

  Frequency-difference (FD) electrical impedance tomography (EIT) using a weighted voltage difference has recently been proposed for imaging haemorrhagic stroke, abdominal bleeding and tumors. Although its feasibility was demonstrated through two-dimensional numerical simulations and phantom experimen... [more] Frequency-difference (FD) electrical impedance tomography (EIT) using a weighted voltage difference has recently been proposed for imaging haemorrhagic stroke, abdominal bleeding and tumors. Although its feasibility was demonstrated through two-dimensional numerical simulations and phantom experiments, we should validate the method in three-dimensional imaging objects. At the same time, we need to investigate its robustness against geometrical modeling errors in boundary shapes and electrode positions. We performed a validation study of the weighted FD method through three-dimensional numerical simulations and phantom experiments. Adopting hemispherical models and phantoms whose admittivity distributions change with frequency, we investigated the performance of the method to detect an anomaly. We found that the simple FD method fails to detect the anomaly, whereas reconstructed images using the weighted FD method clearly visualize the anomaly. The weighted FD method is robust against modeling errors of boundary-shape deformations and displaced electrode positions. We also found that the method is capable of detecting an anomaly surrounded by a shell-shaped obstacle simulating the skull. We propose the weighted FD method for future studies of animal and human experiments.

• Numerical simulation of frequency-difference EIT using multi-shell concentric spherical head model

  Sujin Ahn, Sung Chan Jun, Eung Je Woo

  Noninvasive Functional Source Imaging of the Brain and Heart & 2011 8th International Conference on Bioelectromagnetism (NFSI & ICBEM), 2011 8th International Symposium on; 06/2011

  We present numerical simulations of frequency-difference electrical impedance tomography (EIT) for imaging haemorrhagic stroke inside the head adopting a multi-shell concentric spherical head model. We performed a validation study of a weighted frequency-difference method through three-dimensional n... [more] We present numerical simulations of frequency-difference electrical impedance tomography (EIT) for imaging haemorrhagic stroke inside the head adopting a multi-shell concentric spherical head model. We performed a validation study of a weighted frequency-difference method through three-dimensional numerical simulations. Adopting the multi-shell model with a frequency-dependent admittivity distribution, we investigated the performance of the method to detect a blood anomaly. We found that the simple frequency-difference method fails to detect the anomaly whereas reconstructed images using the weighted frequency-difference method clearly visualize the anomaly. We found that the weighted frequency-difference method is robust against modeling errors since it is capable of detecting an anomaly surrounded by shell-shaped obstacles simulating the internal structures of the head. We propose the weighted frequency-difference method for future studies of animal and human experiments.

• How reducing model mismatch is beneficial to EEG source localization: Simulation study

  Jun Hee Hong, Donghyeon Kim, Sung Chan Jun

  Noninvasive Functional Source Imaging of the Brain and Heart & 2011 8th International Conference on Bioelectromagnetism (NFSI & ICBEM), 2011 8th International Symposium on; 06/2011

  Forward modeling errors as well as measurement noise at sensor surface are propagated into errors in EEG source localization. In order to reduce forward modeling errors, tremendous effort has gone in to estimate real head models as accurately as possible, thereby, yielding more accurate source local... [more] Forward modeling errors as well as measurement noise at sensor surface are propagated into errors in EEG source localization. In order to reduce forward modeling errors, tremendous effort has gone in to estimate real head models as accurately as possible, thereby, yielding more accurate source localization. Additionally, de-noising approaches have been developed to reduce the effect of noise. However, both noise and model mismatch are essentially unavoidable. Typically, the noise level depends on the EEG data to be analyzed. Unaveraged data has a substantially higher noise level than averaged data. For the given noise level of EEG data, how is reducing model mismatch beneficial to EEG source localization? In this work, we attempt to answer this question through an intensive simulation study. Three-shell (representing scalp, skull and brain) concentric spherical head models, with meshes of different fineness are generated. Assuming that the finest mesh model has no modeling errors, about 60,000 single dipole problems are generated on it. Then they are localized on several coarser models using the beamforming technique. Homogeneous conductivity values are assigned for each shell and the finite element method (FEM) is applied for forward computation. Finally, averaged localization error distribution is obtained over signal-to-noise ratios and over different mesh models to see the modeling error effects. It is found that reducing modeling errors has substantial gain in localization, but the gain is marginal after the modeling error is less than a particular value.

• Motor imagery based BCI classification via sparse representation of EEG signals

  Younghak Shin, Seungchan Lee, Minkyu Ahn, Sung Chan Jun, Heung-No Lee

  Noninvasive Functional Source Imaging of the Brain and Heart & 2011 8th International Conference on Bioelectromagnetism (NFSI & ICBEM), 2011 8th International Symposium on; 06/2011

  Electroencephalogram (EEG) based brain-computer interface (BCI) provides a new communication and control channel for people with severe motor disabilities. Motor imagery based sensorimotor rhythm (SMR) analysis is one of the widely used methods in the BCI field. However, these motor imagery signals ... [more] Electroencephalogram (EEG) based brain-computer interface (BCI) provides a new communication and control channel for people with severe motor disabilities. Motor imagery based sensorimotor rhythm (SMR) analysis is one of the widely used methods in the BCI field. However, these motor imagery signals are very noisy and strongly depends on subjects. Therefore, it is difficult to classify them and thus more powerful classification methods are needed. In this paper, we propose a new classification method based on sparse representation of EEG signals and ell-1 minimization. Using Mu and/or Beta rhythm as EEG features, we evaluate the performance of the proposed method with four data sets. Moreover, we make performance comparison with the linear discriminant analysis (LDA), another classification method. From the results, our proposed method shows the better classification accuracy.

• Calibration Time Reduction through Source Imaging in Brain Computer Interface (BCI).

  Minkyu Ahn, Hohyun Cho, Sung Chan Jun

  HCI International 2011 - Posters' Extended Abstracts - International Conference, HCI International 2011, Orlando, FL, USA, July 9-14, 2011, Proceedings, Part II; 01/2011
• 5.74

  Impact points

  Bayesian brain source imaging based on combined MEG/EEG and fMRI using MCMC.

  Sung C Jun, John S George, Woohan Kim, Juliana Paré-Blagoev, Sergey Plis, Doug M Ranken, David M Schmidt

  NeuroImage. 06/2008; 40(4):1581-94.

  A number of brain imaging techniques have been developed in order to investigate brain function and to develop diagnostic tools for various brain disorders. Each modality has strengths as well as weaknesses compared to the others. Recent work has explored how multiple modalities can be integrated ef... [more] A number of brain imaging techniques have been developed in order to investigate brain function and to develop diagnostic tools for various brain disorders. Each modality has strengths as well as weaknesses compared to the others. Recent work has explored how multiple modalities can be integrated effectively so that they complement one another while maintaining their individual strengths. Bayesian inference employing Markov Chain Monte Carlo (MCMC) techniques provides a straightforward way to combine disparate forms of information while dealing with the uncertainty in each. In this paper we introduce methods of Bayesian inference as a way to integrate different forms of brain imaging data in a probabilistic framework. We formulate Bayesian integration of magnetoencephalography (MEG) data and functional magnetic resonance imaging (fMRI) data by incorporating fMRI data into a spatial prior. The usefulness and feasibility of the method are verified through testing with both simulated and empirical data.

  Download

• Source Inversion Technique using Bayesian Inference: Combined MEG/fMRI

  S C Jun, J S George, W. Kim, S. M. Plis, D M Ranken, D M Schmidt

  Noninvasive Functional Source Imaging of the Brain and Heart and the International Conference on Functional Biomedical Imaging, 2007. NFSI-ICFBI 2007. Joint Meeting of the 6th International Symposium on; 11/2007

  As a way to integrate multi-modal brain imaging data in the Bayesian frame, we propose a spatiotemporal Bayesian inference multi-dipole analysis for MEG and fMRI data. We formulate a Bayesian integration of MEG and fMRI data, and its usefulness and feasibility are verified through testing simulated ... [more] As a way to integrate multi-modal brain imaging data in the Bayesian frame, we propose a spatiotemporal Bayesian inference multi-dipole analysis for MEG and fMRI data. We formulate a Bayesian integration of MEG and fMRI data, and its usefulness and feasibility are verified through testing simulated data.
• 2.78

  Impact points

  Probabilistic forward model for electroencephalography source analysis.

  Sergey M Plis, John S George, Sung C Jun, Doug M Ranken, Petr L Volegov, David M Schmidt

  Physics in medicine and biology. 10/2007; 52(17):5309-27.

  Source localization by electroencephalography (EEG) requires an accurate model of head geometry and tissue conductivity. The estimation of source time courses from EEG or from EEG in conjunction with magnetoencephalography (MEG) requires a forward model consistent with true activity for the best out... [more] Source localization by electroencephalography (EEG) requires an accurate model of head geometry and tissue conductivity. The estimation of source time courses from EEG or from EEG in conjunction with magnetoencephalography (MEG) requires a forward model consistent with true activity for the best outcome. Although MRI provides an excellent description of soft tissue anatomy, a high resolution model of the skull (the dominant resistive component of the head) requires CT, which is not justified for routine physiological studies. Although a number of techniques have been employed to estimate tissue conductivity, no present techniques provide the noninvasive 3D tomographic mapping of conductivity that would be desirable. We introduce a formalism for probabilistic forward modeling that allows the propagation of uncertainties in model parameters into possible errors in source localization. We consider uncertainties in the conductivity profile of the skull, but the approach is general and can be extended to other kinds of uncertainties in the forward model. We and others have previously suggested the possibility of extracting conductivity of the skull from measured electroencephalography data by simultaneously optimizing over dipole parameters and the conductivity values required by the forward model. Using Cramer-Rao bounds, we demonstrate that this approach does not improve localization results nor does it produce reliable conductivity estimates. We conclude that the conductivity of the skull has to be either accurately measured by an independent technique, or that the uncertainties in the conductivity values should be reflected in uncertainty in the source location estimates.
• 2.40

  Impact points

  Modeling spatiotemporal covariance for magnetoencephalography or electroencephalography source analysis.

  Sergey M Plis, J S George, S C Jun, J Paré-Blagoev, D M Ranken, C C Wood, D M Schmidt

  Physical review. E, Statistical, nonlinear, and soft matter physics. 02/2007; 75(1 Pt 1):011928.

  We propose a new model to approximate spatiotemporal noise covariance for use in neural electromagnetic source analysis, which better captures temporal variability in background activity. As with other existing formalisms, our model employs a Kronecker product of matrices representing temporal and s... [more] We propose a new model to approximate spatiotemporal noise covariance for use in neural electromagnetic source analysis, which better captures temporal variability in background activity. As with other existing formalisms, our model employs a Kronecker product of matrices representing temporal and spatial covariance. In our model, spatial components are allowed to have differing temporal covariances. Variability is represented as a series of Kronecker products of spatial component covariances and corresponding temporal covariances. Unlike previous attempts to model covariance through a sum of Kronecker products, our model is designed to have a computationally manageable inverse. Despite increased descriptive power, inversion of the model is fast, making it useful in source analysis. We have explored two versions of the model. One is estimated based on the assumption that spatial components of background noise have uncorrelated time courses. Another version, which gives closer approximation, is based on the assumption that time courses are statistically independent. The accuracy of the structural approximation is compared to an existing model, based on a single Kronecker product, using both Frobenius norm of the difference between spatiotemporal sample covariance and a model, and scatter plots. Performance of ours and previous models is compared in source analysis of a large number of single dipole problems with simulated time courses and with background from authentic magnetoencephalography data.
• 2.78

  Impact points

  Spatiotemporal noise covariance estimation from limited empirical magnetoencephalographic data.

  Sung C Jun, Sergey M Plis, Doug M Ranken, David M Schmidt

  Physics in medicine and biology. 12/2006; 51(21):5549-64.

  The performance of parametric magnetoencephalography (MEG) and electroencephalography (EEG) source localization approaches can be degraded by the use of poor background noise covariance estimates. In general, estimation of the noise covariance for spatiotemporal analysis is difficult mainly due to t... [more] The performance of parametric magnetoencephalography (MEG) and electroencephalography (EEG) source localization approaches can be degraded by the use of poor background noise covariance estimates. In general, estimation of the noise covariance for spatiotemporal analysis is difficult mainly due to the limited noise information available. Furthermore, its estimation requires a large amount of storage and a one-time but very large (and sometimes intractable) calculation or its inverse. To overcome these difficulties, noise covariance models consisting of one pair or a sum of multi-pairs of Kronecker products of spatial covariance and temporal covariance have been proposed. However, these approaches cannot be applied when the noise information is very limited, i.e., the amount of noise information is less than the degrees of freedom of the noise covariance models. A common example of this is when only averaged noise data are available for a limited prestimulus region (typically at most a few hundred milliseconds duration). For such cases, a diagonal spatiotemporal noise covariance model consisting of sensor variances with no spatial or temporal correlation has been the common choice for spatiotemporal analysis. In this work, we propose a different noise covariance model which consists of diagonal spatial noise covariance and Toeplitz temporal noise covariance. It can easily be estimated from limited noise information, and no time-consuming optimization and data-processing are required. Thus, it can be used as an alternative choice when one-pair or multi-pair noise covariance models cannot be estimated due to lack of noise information. To verify its capability we used Bayesian inference dipole analysis and a number of simulated and empirical datasets. We compared this covariance model with other existing covariance models such as conventional diagonal covariance, one-pair and multi-pair noise covariance models, when noise information is sufficient to estimate them. We found that our proposed noise covariance model yields better localization performance than a diagonal noise covariance, while it performs slightly worse than one-pair or multi-pair noise covariance models - although these require much more noise information. Finally, we present some localization results on median nerve stimulus empirical MEG data for our proposed noise covariance model.

• A generalized spatiotemporal covariance model for stationary background in analysis of MEG data

  S.M. Plis, D.M. Schmidt, S.C. Jun, D.M. Ranken

  Engineering in Medicine and Biology Society, 2006. EMBS '06. 28th Annual International Conference of the IEEE; 10/2006

  Using a noise covariance model based on a single Kronecker product of spatial and temporal covariance in the spatiotemporal analysis of MEG data was demonstrated to provide improvement in the results over that of the commonly used diagonal noise covariance model. In this paper we present a model tha... [more] Using a noise covariance model based on a single Kronecker product of spatial and temporal covariance in the spatiotemporal analysis of MEG data was demonstrated to provide improvement in the results over that of the commonly used diagonal noise covariance model. In this paper we present a model that is a generalization of all of the above models. It describes models based on a single Kronecker product of spatial and temporal covariance as well as more complicated multi-pair models together with any intermediate form expressed as a sum of Kronecker products of spatial component matrices of reduced rank and their corresponding temporal covariance matrices. The model provides a framework for controlling the tradeoff between the described complexity of the background and computational demand for the analysis using this model. Ways to estimate the value of the parameter controlling this tradeoff are also discussed
• 2.78

  Impact points

  Improving source detection and separation in a spatiotemporal Bayesian inference dipole analysis.

  Sung C Jun, John S George, Sergey M Plis, Doug M Ranken, David M Schmidt, C C Wood

  Physics in medicine and biology. 06/2006; 51(10):2395-414.

  Most existing spatiotemporal multi-dipole approaches for MEG/EEG source localization assume that the dipoles are active for the full time range being analysed. If the actual time range of activity of sources is significantly shorter than the time range being analysed, the detectability, localization... [more] Most existing spatiotemporal multi-dipole approaches for MEG/EEG source localization assume that the dipoles are active for the full time range being analysed. If the actual time range of activity of sources is significantly shorter than the time range being analysed, the detectability, localization and time-course determination of such sources may be adversely affected, especially for weak sources. In order to improve detectability and reconstruction of such sources, it is natural to add active time range information (starting time point and ending time point of source activation) for each candidate source as unknown parameters in the analysis. However, this adds additional nonlinear free parameters that could burden the analysis and could be unfeasible for some methods. Recently, we described a spatiotemporal Bayesian inference multi-dipole analysis for the MEG/EEG inverse problem. This approach treated the number of dipoles as a free parameter, produced realistic uncertainty estimates using a Markov chain Monte Carlo numerical sampling of the posterior distribution and included a method to reduce the unwanted effects of local minima. In this paper, our spatiotemporal Bayesian inference multi-dipole analysis is extended to incorporate active time range parameters of starting and stopping time points. The properties of this analysis in comparison to the previous one without active time range parameters are demonstrated through extensive studies using both simulated and empirical MEG data.

• A generalized spatiotemporal covariance model for stationary background in analysis of MEG data.

  S. M. Plis, D M Schmidt, S C Jun, D M Ranken

  Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference. 02/2006; 1:3680-3.

  Using a noise covariance model based on a single Kronecker product of spatial and temporal covariance in the spatiotemporal analysis of MEG data was demonstrated to provide improvement in the results over that of the commonly used diagonal noise covariance model. In this paper we present a model tha... [more] Using a noise covariance model based on a single Kronecker product of spatial and temporal covariance in the spatiotemporal analysis of MEG data was demonstrated to provide improvement in the results over that of the commonly used diagonal noise covariance model. In this paper we present a model that is a generalization of all of the above models. It describes models based on a single Kronecker product of spatial and temporal covariance as well as more complicated multi-pair models together with any intermediate form expressed as a sum of Kronecker products of spatial component matrices of reduced rank and their corresponding temporal covariance matrices. The model provides a framework for controlling the tradeoff between the described complexity of the background and computational demand for the analysis using this model. Ways to estimate the value of the parameter controlling this tradeoff are also discussed.
• 5.74

  Impact points

  Spatiotemporal Bayesian inference dipole analysis for MEG neuroimaging data.

  Sung C Jun, John S George, Juliana Paré-Blagoev, Sergey M Plis, Doug M Ranken, David M Schmidt, C C Wood

  NeuroImage. 11/2005; 28(1):84-98.

  Recently, we described a Bayesian inference approach to the MEG/EEG inverse problem that used numerical techniques to estimate the full posterior probability distributions of likely solutions upon which all inferences were based [Schmidt, D.M., George, J.S., Wood, C.C., 1999. Bayesian inference appl... [more] Recently, we described a Bayesian inference approach to the MEG/EEG inverse problem that used numerical techniques to estimate the full posterior probability distributions of likely solutions upon which all inferences were based [Schmidt, D.M., George, J.S., Wood, C.C., 1999. Bayesian inference applied to the electromagnetic inverse problem. Human Brain Mapping 7, 195; Schmidt, D.M., George, J.S., Ranken, D.M., Wood, C.C., 2001. Spatial-temporal bayesian inference for MEG/EEG. In: Nenonen, J., Ilmoniemi, R. J., Katila, T. (Eds.), Biomag 2000: 12th International Conference on Biomagnetism. Espoo, Norway, p. 671]. Schmidt et al. (1999) focused on the analysis of data at a single point in time employing an extended region source model. They subsequently extended their work to a spatiotemporal Bayesian inference analysis of the full spatiotemporal MEG/EEG data set. Here, we formulate spatiotemporal Bayesian inference analysis using a multi-dipole model of neural activity. This approach is faster than the extended region model, does not require use of the subject's anatomical information, does not require prior determination of the number of dipoles, and yields quantitative probabilistic inferences. In addition, we have incorporated the ability to handle much more complex and realistic estimates of the background noise, which may be represented as a sum of Kronecker products of temporal and spatial noise covariance components. This reduces the effects of undermodeling noise. In order to reduce the rigidity of the multi-dipole formulation which commonly causes problems due to multiple local minima, we treat the given covariance of the background as uncertain and marginalize over it in the analysis. Markov Chain Monte Carlo (MCMC) was used to sample the many possible likely solutions. The spatiotemporal Bayesian dipole analysis is demonstrated using simulated and empirical whole-head MEG data.

• Modeling spatiotemporal noise covariance for MEG/EEG source analysis

  S. M. Plis, J S George, S C Jun, J. Pare-Blagoev, D M Ranken, D M Schmidt, C C Wood

  04/2005;

  We propose a new model for approximating spatiotemporal noise covariance for use in MEG/EEG source analysis. Our model is an extension of an existing model [1,2] that uses a single Kronecker product of a pair of matrices - temporal and spatial covariance; we employ a series of Kronecker products in ... [more] We propose a new model for approximating spatiotemporal noise covariance for use in MEG/EEG source analysis. Our model is an extension of an existing model [1,2] that uses a single Kronecker product of a pair of matrices - temporal and spatial covariance; we employ a series of Kronecker products in order to construct a better approximation of the full covariance. In contrast to the single-pair model that assumes the same temporal structure for all spatial components, the proposed model allows for distinct, independent time courses at each spatial component. This model better describes spatially and temporally correlated background activity. At the same time, inversion of the model is fast which makes it useful in the inverse analysis. We have explored two versions of the model. One is based on orthogonal spatial components of the background. The other, more general model, is based on independent spatial components. Performance of the new and previous models is compared in inverse solutions to a large number of single dipole problems with simulated time courses and background from authentic MEG data.
• 6.26

  Impact points

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

  Sung Chan Jun, Barak A. Pearlmutter

  Human brain mapping. 02/2005; 24(1):21-34.

  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 previo... [more] 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.

• Subject-Independent Magnetoencephalographic Source Localization by a Multilayer Perceptron

  Sung C. Jun, Barak A. Pearlmutter

  03/2004;

  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 previo... [more] 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.
• 2.15

  Impact points

  MEG source localization using an MLP with a distributed output representation.

  Sung Chan Jun, Barak A. Pearlmutter, Guido Nolte

  IEEE transactions on bio-medical engineering. 07/2003; 50(6):786-9.

  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 amplitude... [more] 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.