Kwang-Ok An

Assessment criteria for MEG/EEG cortical patch tests

To validate newly developed methods or implemented software for magnetoencephalography/electroencephalography (MEG/EEG) source localization problems, many researchers have used human skull phantom experiments or artificially constructed forward data sets. Between the two methods, the use of an artificial data set constructed with forward calculation attains superiority over the use of a human skull phantom in that it is simple to implement, adjust and control various conditions. Nowadays, for the forward calculation, especially for the cortically distributed source models, generating artificial activation patches on a brain cortical surface has been popularized instead of activating some point dipole sources. However, no well-established assessment criterion to validate the reconstructed results quantitatively has yet been introduced. In this paper, we suggest some assessment criteria to compare and validate the various MEG/EEG source localization techniques or implemented software applied to the cortically distributed source model. Four different criteria can be used to measure accuracy, degrees of focalization, noise-robustness, existence of spurious sources and so on. To verify the usefulness of the proposed criteria, four different results from two different noise conditions and two different reconstruction techniques were compared for several patches. The simulated results show that the new criteria can provide us with a reliable index to validate the MEG/EEG source localization techniques.

Precise Estimation of Brain Electrical Sources Using Anatomically Constrained Area Source (ACAS) Localization

This paper proposes a new parametric source model to precisely estimate brain electrical source distribution from electroencephalogram (EEG) or magnetoencephalogram recordings. In the proposed approach, only four free parameters need to be evaluated for each anatomically constrained area source to express both spatial locations and extensions of brain electrical activities on cortical surface. The feasibility of the proposed model is verified through the applications to realistic EEG simulations and epilepsy patients data

Reconstruction of Continuous and Focalized Brain Functional Source Images from Electroencephalography

In this paper, a new hybrid approach to reconstruct more accurate brain functional source images from electroencephalography is proposed. The proposed approach combines extended source model and focal underdetermined system solution algorithm. Feasibility studies with realistic simulation data and the epilepsy patients data demonstrate that continuous, as well as focalized, brain electrical source images can be reconstructed utilizing the proposed approach

Clinical feasibility of brain-computer interface based on steady-state visual evoked potential in patients with locked-in syndrome: Case studies.

Although the feasibility of brain-computer interface (BCI) systems based on steady-state visual evoked potential (SSVEP) has been extensively investigated, only a few studies have evaluated its clinical feasibility in patients with locked-in syndrome (LIS), who are the main targets of BCI technology. The main objective of this case report was to share our experiences of SSVEP-based BCI experiments involving five patients with LIS, thereby providing researchers with useful information that can potentially help them to design BCI experiments for patients with LIS. In our experiments, a four-class online SSVEP-based BCI system was implemented and applied to four of five patients repeatedly on multiple days to investigate its test-retest reliability. In the last experiments with two of the four patients, the practical usability of our BCI system was tested using a questionnaire survey. All five patients showed clear and distinct SSVEP responses at all four fundamental stimulation frequencies (6, 6.66, 7.5, 10 Hz), and responses at harmonic frequencies were also observed in three patients. Mean classification accuracy was 76.99% (chance level = 25%). The test-retest reliability experiments demonstrated stable performance of our BCI system over different days even when the initial experimental settings (e.g., electrode configuration, fixation time, visual angle) used in the first experiment were used without significant modifications. Our results suggest that SSVEP-based BCI paradigms might be successfully used to implement clinically feasible BCI systems for severely paralyzed patients.

Magnetoencephalography source localization using improved simplex method

Nelder–Mead downhill simplex method, a kind of deterministic optimization algorithms, has been used extensively for magnetoencephalography (MEG) dipolar source localization problems because it does not require any functional differentiation. Like many other deterministic algorithms, however, it can be easily trapped in local optima when being applied to complex inverse problems with multiple simultaneous sources. In the present study, some modifications have been made to improve its capability of finding global optima. Those include (1) constructing an initial simplex based upon sensitivity of variables and (2) introducing a shaking technique based on polynomial interpolation. The efficiency of the proposed method was tested using analytical test functions and simulated MEG data. The simulation results demonstrate that the improved downhill simplex method can result in more reliable inverse solutions than the conventional one.

Enhancing accuracy in magneto-and electroencephalography focal source localization

This paper investigates the origins of localization errors when cortically distributed source approach is applied to magnetoencephalography (MEG) and electroencephalography (EEG) focal source reconstruction problems. It was revealed from simulations that the errors mainly stem from 1)Asymmetric distribution of crosstalk and 2)Large noises at some time slices. To enhance the localization accuracy, a new algorithm named cortical surface scanning (CSS) method is presented. The proposed algorithm was applied to realistic simulations and more accurate and robust source estimates could be obtained

Novel multidipole searching technique for magnetoencephalography source localization

A multidipole searching technique using a moving dipole concept is proposed for magnetoencephalography (MEG) source localization. The proposed method can be a promising method to resolve a critical disadvantage of conventional equivalent current dipole (ECD) method-the number of current dipoles should be given a priori. From several simulation results, the accuracy and effectiveness of the method will be verified.

Improved magnetoencephalography source reconstruction considering anatomical connectivity of cortical sources

In this paper, an improved magnetoencephalography (MEG) source reconstruction technique considering anatomical connectivity of cortical sources is proposed. The anatomical connectivity information was taken into account by calculating three-dimensional geodesic distance between neighboring sources, and then the resultant inverse solutions were compared with those of other cases:1)Inverse estimate without connectivity information; 2)Use of Euclidean distance instead of geodesic distance. The proposed technique was applied to realistic simulations for a real brain anatomy, and the results showed that estimated sources can be smoother and more accurate by using the anatomical connectivity information

Estimation of Solution Accuracy in Magnetoencephalography by using the Condition Number from a Leadfield Matrix

This paper is a report on an investigation of the relationship between the condition number of a leadfield matrix of magnetoencephalography (MEG) and source reconstruction accuracy. We suggest the use of the condition number as a promising a priori accuracy estimator of neuromagnetic inverse problem (NIP) since various simulation studies demonstrated that the reconstructed source distribution has the wider region and contains the more spurious sources if a leadfield matrix has the higher condition number. The simulation results also verified that the accuracy estimator based upon the condition number can explain well-known phenomena of NIP, which suggests a potential application of the proposed concept.

Enhancement of calculation time and accuracy in MEG inverse problem using simplex method

In magnetoencephalography (MEG) source localization, simplex method is widely used because it is simple and does not need the derivatives of the function. However, the simplex method is very sensitive to the choice of initial points. Thus, in this paper, a method of choosing the good initial starting points is proposed. First, the brain is divided into several partitions and the value of center of each partition is evaluated. Then, the point which has the lowest value is determined to be the initial point. In addition, we estimate the calculation time and the accuracy of the source localization according to increasing the number of partitions.